One document matched: draft-ietf-dccp-tfrc-voip-06.ps

%!PS-Adobe-3.0
%%Creator: groff version 1.18.1
%%CreationDate: Sun Oct 22 22:00:59 2006
%%DocumentNeededResources: font Courier-Bold
%%+ font Times-Bold
%%+ font Times-Roman
%%+ font Times-Italic
%%+ font Courier
%%DocumentSuppliedResources: procset grops 1.18 1
%%Pages: 36
%%PageOrder: Ascend
%%Orientation: Portrait
%%EndComments
%%BeginProlog
%%BeginResource: procset grops 1.18 1
/setpacking where{
pop
currentpacking
true setpacking
}if
/grops 120 dict dup begin
/SC 32 def
/A/show load def
/B{0 SC 3 -1 roll widthshow}bind def
/C{0 exch ashow}bind def
/D{0 exch 0 SC 5 2 roll awidthshow}bind def
/E{0 rmoveto show}bind def
/F{0 rmoveto 0 SC 3 -1 roll widthshow}bind def
/G{0 rmoveto 0 exch ashow}bind def
/H{0 rmoveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/I{0 exch rmoveto show}bind def
/J{0 exch rmoveto 0 SC 3 -1 roll widthshow}bind def
/K{0 exch rmoveto 0 exch ashow}bind def
/L{0 exch rmoveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/M{rmoveto show}bind def
/N{rmoveto 0 SC 3 -1 roll widthshow}bind def
/O{rmoveto 0 exch ashow}bind def
/P{rmoveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/Q{moveto show}bind def
/R{moveto 0 SC 3 -1 roll widthshow}bind def
/S{moveto 0 exch ashow}bind def
/T{moveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/SF{
findfont exch
[exch dup 0 exch 0 exch neg 0 0]makefont
dup setfont
[exch/setfont cvx]cvx bind def
}bind def
/MF{
findfont
[5 2 roll
0 3 1 roll
neg 0 0]makefont
dup setfont
[exch/setfont cvx]cvx bind def
}bind def
/level0 0 def
/RES 0 def
/PL 0 def
/LS 0 def
/MANUAL{
statusdict begin/manualfeed true store end
}bind def
/PLG{
gsave newpath clippath pathbbox grestore
exch pop add exch pop
}bind def
/BP{10 -52 translate
/level0 save def
1 setlinecap
1 setlinejoin
72 RES div dup scale
LS{
90 rotate
}{
0 PL translate
}ifelse
1 -1 scale
}bind def
/EP{
level0 restore
showpage
}bind def
/DA{
newpath arcn stroke
}bind def
/SN{
transform
.25 sub exch .25 sub exch
round .25 add exch round .25 add exch
itransform
}bind def
/DL{
SN
moveto
SN
lineto stroke
}bind def
/DC{
newpath 0 360 arc closepath
}bind def
/TM matrix def
/DE{
TM currentmatrix pop
translate scale newpath 0 0 .5 0 360 arc closepath
TM setmatrix
}bind def
/RC/rcurveto load def
/RL/rlineto load def
/ST/stroke load def
/MT/moveto load def
/CL/closepath load def
/Fr{
setrgbcolor fill
}bind def
/Fk{
setcmykcolor fill
}bind def
/Fg{
setgray fill
}bind def
/FL/fill load def
/LW/setlinewidth load def
/Cr/setrgbcolor load def
/Ck/setcmykcolor load def
/Cg/setgray load def
/RE{
findfont
dup maxlength 1 index/FontName known not{1 add}if dict begin
{
1 index/FID ne{def}{pop pop}ifelse
}forall
/Encoding exch def
dup/FontName exch def
currentdict end definefont pop
}bind def
/DEFS 0 def
/EBEGIN{
moveto
DEFS begin
}bind def
/EEND/end load def
/CNT 0 def
/level1 0 def
/PBEGIN{
/level1 save def
translate
div 3 1 roll div exch scale
neg exch neg exch translate
0 setgray
0 setlinecap
1 setlinewidth
0 setlinejoin
10 setmiterlimit
[]0 setdash
/setstrokeadjust where{
pop
false setstrokeadjust
}if
/setoverprint where{
pop
false setoverprint
}if
newpath
/CNT countdictstack def
userdict begin
/showpage{}def
}bind def
/PEND{
clear
countdictstack CNT sub{end}repeat
level1 restore
}bind def
end def
/setpacking where{
pop
setpacking
}if
%%EndResource
%%IncludeResource: font Courier-Bold
%%IncludeResource: font Times-Bold
%%IncludeResource: font Times-Roman
%%IncludeResource: font Times-Italic
%%IncludeResource: font Courier
grops begin/DEFS 1 dict def DEFS begin/u{.001 mul}bind def end/RES 72
def/PL 841.89 def/LS false def/ENC0[/asciicircum/asciitilde/Scaron
/Zcaron/scaron/zcaron/Ydieresis/trademark/quotesingle/Euro/.notdef
/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef
/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef
/.notdef/.notdef/.notdef/space/exclam/quotedbl/numbersign/dollar/percent
/ampersand/quoteright/parenleft/parenright/asterisk/plus/comma/hyphen
/period/slash/zero/one/two/three/four/five/six/seven/eight/nine/colon
/semicolon/less/equal/greater/question/at/A/B/C/D/E/F/G/H/I/J/K/L/M/N/O
/P/Q/R/S/T/U/V/W/X/Y/Z/bracketleft/backslash/bracketright/circumflex
/underscore/quoteleft/a/b/c/d/e/f/g/h/i/j/k/l/m/n/o/p/q/r/s/t/u/v/w/x/y
/z/braceleft/bar/braceright/tilde/.notdef/quotesinglbase/guillemotleft
/guillemotright/bullet/florin/fraction/perthousand/dagger/daggerdbl
/endash/emdash/ff/fi/fl/ffi/ffl/dotlessi/dotlessj/grave/hungarumlaut
/dotaccent/breve/caron/ring/ogonek/quotedblleft/quotedblright/oe/lslash
/quotedblbase/OE/Lslash/.notdef/exclamdown/cent/sterling/currency/yen
/brokenbar/section/dieresis/copyright/ordfeminine/guilsinglleft
/logicalnot/minus/registered/macron/degree/plusminus/twosuperior
/threesuperior/acute/mu/paragraph/periodcentered/cedilla/onesuperior
/ordmasculine/guilsinglright/onequarter/onehalf/threequarters
/questiondown/Agrave/Aacute/Acircumflex/Atilde/Adieresis/Aring/AE
/Ccedilla/Egrave/Eacute/Ecircumflex/Edieresis/Igrave/Iacute/Icircumflex
/Idieresis/Eth/Ntilde/Ograve/Oacute/Ocircumflex/Otilde/Odieresis
/multiply/Oslash/Ugrave/Uacute/Ucircumflex/Udieresis/Yacute/Thorn
/germandbls/agrave/aacute/acircumflex/atilde/adieresis/aring/ae/ccedilla
/egrave/eacute/ecircumflex/edieresis/igrave/iacute/icircumflex/idieresis
/eth/ntilde/ograve/oacute/ocircumflex/otilde/odieresis/divide/oslash
/ugrave/uacute/ucircumflex/udieresis/yacute/thorn/ydieresis]def
/Courier@0 ENC0/Courier RE/Times-Italic@0 ENC0/Times-Italic RE
/Times-Roman@0 ENC0/Times-Roman RE/Times-Bold@0 ENC0/Times-Bold RE
/Courier-Bold@0 ENC0/Courier-Bold RE
%%EndProlog
%%Page: 1 1
%%BeginPageSetup
BP
%%EndPageSetup
/F0 10/Courier-Bold@0 SF(Internet Engineering Task Force)72 85 Q
(Sally Floyd)179.999 E 317.999(INTERNET-DRAFT ICIR)72 98 R 167.999
(draft-ietf-dccp-tfrc-voip-06.ps Eddie)72 111 R(Kohler)6 E
(Expires: April 2007)72 124 Q(UCLA)293.999 E(22 October 2006)413.999 137
Q/F1 14/Times-Bold@0 SF(TCP Friendly Rate Contr)178.086 162 Q
(ol \(TFRC\):)-.252 E(the Small-P)198.379 175 Q(ack)-.14 E(et \(SP\) V)
-.14 E(ariant)-1.288 E/F2 11/Times-Bold@0 SF(Status of this Memo)72 207
Q/F3 11/Times-Roman@0 SF
(By submitting this Internet-Draft, each author represents that an)100.8
223.6 Q 2.75(ya)-.165 G(pplicable patent or other)-2.75 E
(IPR claims of which he or she is a)100.8 236.6 Q -.11(wa)-.165 G(re ha)
.11 E .33 -.165(ve b)-.22 H(een or will be disclosed, and an).165 E 2.75
(yo)-.165 G 2.75(fw)-2.75 G(hich he)-2.75 E(or she becomes a)100.8 249.6
Q -.11(wa)-.165 G
(re will be disclosed, in accordance with Section 6 of BCP 79.).11 E
(Internet-Drafts are w)100.8 266.2 Q
(orking documents of the Internet Engineering T)-.11 E(ask F)-.88 E
(orce \(IETF\), its)-.165 E(areas, and its w)100.8 279.2 Q
(orking groups.)-.11 E(Note that other groups may also distrib)5.5 E
(ute w)-.22 E(orking)-.11 E(documents as Internet-Drafts.)100.8 292.2 Q
(Internet-Drafts are draft documents v)100.8 308.8 Q
(alid for a maximum of six months and may be)-.275 E
(updated, replaced, or obsoleted by other documents at an)100.8 321.8 Q
2.75(yt)-.165 G 2.75(ime. It)-2.75 F(is inappropriate to use)2.75 E
(Internet-Drafts as reference material or to cite them other than as "w)
100.8 334.8 Q(ork in progress.")-.11 E
(The list of current Internet-Drafts can be accessed at http://www)100.8
351.4 Q(.ietf.or)-.715 E(g/ietf/1id-)-.198 E(abstracts.txt.)100.8 364.4
Q(The list of Internet-Draft Shado)100.8 381 Q 2.75(wD)-.275 G
(irectories can be accessed at)-2.75 E(http://www)100.8 394 Q(.ietf.or)
-.715 E(g/shado)-.198 E -.715(w.)-.275 G(html.).715 E
(This Internet-Draft will e)100.8 410.6 Q(xpire on April 2007.)-.165 E
F2(Abstract)72 436.6 Q F3
(This document proposes a mechanism for further e)100.8 453.2 Q
(xperimentation, b)-.165 E(ut not for widespread)-.22 E(deplo)100.8
466.2 Q(yment at this time in the global Internet.)-.11 E(TCP-Friendly \
Rate Control \(TFRC\) is a congestion control mechanism for unicast \
\215o)100.8 482.8 Q(ws)-.275 E(operating in a best-ef)100.8 495.8 Q
(fort Internet en)-.275 E(vironment [RFC3448]. TFRC w)-.44 E
(as intended for)-.11 E(applications that use a \214x)100.8 508.8 Q
(ed pack)-.165 E(et size, and w)-.11 E(as designed to be reasonably f)
-.11 E(air when)-.11 E
(competing for bandwidth with TCP connections using the same pack)100.8
521.8 Q(et size.)-.11 E(This)5.5 E(document proposes TFRC-SP)100.8 534.8
Q 2.75(,aS)-1.221 G(mall-P)-2.75 E(ack)-.165 E(et \(SP\) v)-.11 E
(ariant of TFRC, that is designed for)-.275 E
(applications that send small pack)100.8 547.8 Q 2.75(ets. The)-.11 F
(design goal for TFRC-SP is to achie)2.75 E .33 -.165(ve t)-.275 H
(he same).165 E(bandwidth in bps \(bits per second\) as a TCP \215o)
100.8 560.8 Q 2.75(wu)-.275 G(sing pack)-2.75 E
(ets of up to 1500 bytes.)-.11 E(TFRC-SP enforces a minimum interv)100.8
573.8 Q(al of 10 ms between data pack)-.275 E(ets, to pre)-.11 E -.165
(ve)-.275 G(nt a single).165 E<8d6f>100.8 586.8 Q 2.75(wf)-.275 G
(rom sending small pack)-2.75 E(ets arbitrarily frequently)-.11 E(.)
-.715 E(Flo)100.8 603.4 Q(ws using TFRC-SP compete reasonably f)-.275 E
(airly with lar)-.11 E(ge-pack)-.198 E(et TCP and TFRC \215o)-.11 E(ws)
-.275 E(in en)100.8 616.4 Q(vironments where lar)-.44 E(ge-pack)-.198 E
(et \215o)-.11 E(ws and small-pack)-.275 E(et \215o)-.11 E(ws e)-.275 E
(xperience similar pack)-.165 E(et)-.11 E(drop rates.)100.8 629.4 Q(Ho)
5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, i).165 H 2.75(ne).44 G
-.44(nv)-2.75 G(ironments where small-pack).44 E(et \215o)-.11 E(ws e)
-.275 E(xperience lo)-.165 E(wer pack)-.275 E(et)-.11 E
(drop rates than lar)100.8 642.4 Q(ge-pack)-.198 E(et \215o)-.11 E
(ws \(e.g., with Drop-T)-.275 E(ail queues in units of bytes\), TFRC-SP)
-.88 E(can recei)100.8 655.4 Q .33 -.165(ve c)-.275 H
(onsiderably more than its share of the bandwidth.).165 E(Flo)72 733 Q
(yd/K)-.11 E 333.671(ohler [P)-.385 F(age 1])-.165 E 0 Cg EP
%%Page: 2 2
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E .396 -.198(TO B)100.8 85 T
2.75(ED).198 G(ELETED BY THE RFC EDIT)-2.75 E(OR UPON PUBLICA)-.198 E
(TION:)-1.221 E(Changes from draft-ietf-dccp-tfrc-v)103.55 98 Q
(oip-05.txt:)-.22 E 2.75(*F)106.3 111 S(eedback from Gorry F)-2.75 E
(airhurst:)-.165 E 2.75(-S)111.8 124 S
(mall editing changes, rephrasing, and b)-2.75 E(ug \214x)-.22 E(es.)
-.165 E 2.75(-E)111.8 137 S
(xplicitly stated that assuming a 40-byte header)-2.75 E(is OK e)117.3
150 Q -.165(ve)-.275 G 2.75(ni).165 G 2.75(fI)-2.75 G(Pv6 is used.)-2.75
E 2.75(-M)111.8 163 S -.165(ove)-2.75 G 2.75(dm).165 G
(ost of Section 7 on Simulations to)-2.75 E(the appendix.)117.3 176 Q
2.75(-A)111.8 189 S(dded a subsection to Section 8 on "F)-2.75 E
(airness with)-.165 E(dif)117.3 202 Q(ferent pack)-.275 E
(et header sizes")-.11 E 2.75(-A)111.8 215 S
(dd a paragraph about TFRC-SP and TFRC-PS.)-2.75 E 2.75(-As)111.8 228 S
(udden step-change in the pack)-2.75 E(et size? -)-.11 E
(no change made to algorithm.)122.8 241 Q 2.75(-C)111.8 254 S
(lari\214ed te)-2.75 E(xt on impact of a reduced TCP MSS.)-.165 E 2.75
(*F)106.3 267 S(eedback from Lars Eggert:)-2.75 E 2.75(-S)111.8 280 S
(mall editing changes, rephrasing, and b)-2.75 E(ug \214x)-.22 E(es.)
-.165 E 2.75(*F)106.3 293 S(eedback from Mark Handle)-2.75 E(y:)-.165 E
2.75(-N)111.8 306 S 2.75(oc)-2.75 G(hange made to algorithm, b)-2.75 E
(ut added Appendix C)-.22 E(to discuss the issue.)117.3 319 Q 2.75(*F)
106.3 332 S(eedback from Ladan Gharai:)-2.75 E 2.75(-A)111.8 345 S
(dded the restriction that the most recent loss interv)-2.75 E(al)-.275
E(is not included in the calculation of the a)117.3 358 Q -.165(ve)-.22
G(rage).165 E(loss interv)117.3 371 Q(al if the most recent loss interv)
-.275 E(al is short.)-.275 E 2.75(-C)111.8 384 S(orrected use of "pack)
-2.75 E(et" and "se)-.11 E(gment" in a fe)-.165 E 2.75(wp)-.275 G
(laces.)-2.75 E(Changes from draft-ietf-dccp-tfrc-v)103.55 410 Q
(oip-04.txt:)-.22 E 2.75(*A)106.3 423 S(dded tables sho)-2.75 E
(wing the response function for TCP)-.275 E 2.75(,T)-1.221 G(FRC,)-2.75
E(and TFRC-SP)111.8 436 Q 2.75(,f)-1.221 G(or a range of pack)-2.75 E
(et sizes, for both)-.11 E(pack)111.8 449 Q(et and byte drop rates.)-.11
E(In response to feedback)5.5 E(from Magnus W)111.8 462 Q(esterlund.)
-.88 E 2.75(*A)106.3 475 S(long with response function, added that TCP')
-2.75 E 2.75(ss)-.605 G(ending)-2.75 E(rate v)111.8 488 Q
(aries linearly with pack)-.275 E(et size.)-.11 E(From a)5.5 E
(suggestion by Magnus W)111.8 501 Q(esterlund.)-.88 E 2.75(*A)106.3 514
S(dded a sentence saying that TCP has a range of sender)-2.75 E
(algorithms for setting the R)111.8 527 Q -.198(TO)-.66 G(.).198 E 2.75
(*D)106.3 540 S(eleted the sentence equating TFRC-SP with TFRC-PS)-2.75
E(referred to in RFC 3448.)111.8 553 Q(From a suggestion)5.5 E
(by Colin Perkins.)111.8 566 Q 2.75(*A)106.3 579 S
(dded that wireless links sometimes are less lik)-2.75 E(ely to)-.11 E
(drop small pack)111.8 592 Q 2.75(ets. Reported)-.11 F
(from Pete Sholander)2.75 E(.)-.605 E 2.75(*A)106.3 605 S
(dded simulations to the end of Section 7.3 comparing)-2.75 E(the ef)
111.8 618 Q(fects of TFRC and of TFRC-SP)-.275 E 2.75(,f)-1.221 G
(or an en)-2.75 E(vironment)-.44 E(with a Drop-T)111.8 631 Q
(ail queue in bytes, sho)-.88 E(wing the possible)-.275 E(ne)111.8 644 Q
-.055(ga)-.165 G(ti).055 E .33 -.165(ve c)-.275 H
(onsequences of TFRC-SP).165 E 5.5(.I)-1.221 G 2.75(nr)-5.5 G
(esponse to email)-2.75 E(from Magnus W)111.8 657 Q(esterlund.)-.88 E
2.75(*A)106.3 670 S(dded an e)-2.75 E(xplanation for the Adapti)-.165 E
.33 -.165(ve R)-.275 H(ED simulation).165 E(with pack)111.8 683 Q
(et drop rates greater than 50%. In response)-.11 E
(to email from Magnus W)111.8 696 Q(esterlund.)-.88 E(Flo)72 733 Q(yd/K)
-.11 E 333.671(ohler [P)-.385 F(age 2])-.165 E 0 Cg EP
%%Page: 3 3
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E 2.75(*A)106.3 85 S
(dded a Conclusions section, with a sentence that a)-2.75 E
(separate document will be used to specify an)111.8 98 Q -.165(ex)111.8
111 S(perimental CCID based on TFRC-SP).165 E 5.5(.I)-1.221 G 2.75(nr)
-5.5 G(esponse)-2.75 E(to feedback during W)111.8 124 Q
(orking Group Last Call.)-.88 E 2.75(*A)106.3 137 S
(dded a paragraph about "Initializing the Loss History)-2.75 E
(after the First Loss Ev)111.8 150 Q(ent" in TFRC-SP)-.165 E(.)-1.221 E
(Changes from draft-ietf-dccp-tfrc-v)103.55 176 Q(oip-03.txt:)-.22 E
2.75(*A)106.3 189 S(dded a paragraph saying that this is intended for)
-2.75 E(Experimental, for further e)111.8 202 Q(xperimentation and not)
-.165 E(for widespread deplo)111.8 215 Q(yment.)-.11 E 2.75(*E)106.3 228
S(diting of abstract so that it still \214ts the 25-line)-2.75 E(limit.)
111.8 241 Q(Changes from draft-ietf-dccp-tfrc-v)103.55 267 Q
(oip-02.txt:)-.22 E 2.75(*C)106.3 280 S(hanged name from "V)-2.75 E
(oIP v)-1.419 E(ariant of TFRC" to "TFRC-SP".)-.275 E 2.75(*A)106.3 293
S(dded Section 4.5 on "The Nominal P)-2.75 E(ack)-.165 E
(et Size", discussing)-.11 E(possible dif)114.55 306 Q(ferences in pack)
-.275 E(et drop rates between small)-.11 E(and lar)114.55 319 Q(ge pack)
-.198 E(ets.)-.11 E 2.75(*A)106.3 332 S(dded te)-2.75 E
(xt to Section 5 on "A Comparison with RFC 3714".)-.165 E 2.75(*A)106.3
345 S(dded te)-2.75 E
(xt to Section 6 on "TFRC-SP with Applications that)-.165 E
(Modify the P)117.3 358 Q(ack)-.165 E(et Size")-.11 E 2.75(*A)106.3 371
S(dded simulations with small-pack)-2.75 E(et TCP \215o)-.11 E(ws.)-.275
E 2.75(*A)106.3 384 S(dded a Security Considerations section.)-2.75 E
2.75(*M)106.3 397 S(inor editing.)-2.75 E
(Changes from draft-ietf-dccp-tfrc-v)103.55 423 Q(oip-01.txt:)-.22 E
2.75(*A)106.3 436 S
(dded modi\214ed algorithm for calculating the loss e)-2.75 E -.165(ve)
-.275 G(nt rate,).165 E(for short interv)117.3 449 Q
(als with multiple pack)-.275 E(et drops.)-.11 E 2.75(*M)106.3 462 S
-.165(ove)-2.75 G 2.75(dF).165 G(aster Restart to a separate document.)
-2.915 E 2.75(*A)106.3 475 S
(dded simulations with a con\214gured byte drop rate.)-2.75 E 2.75(*A)
106.3 488 S(dded man)-2.75 E 2.75(ym)-.165 G
(ore simulations, including Drop-T)-2.75 E(ail with a queue)-.88 E
(in bytes.)111.8 501 Q 2.75(*A)106.3 514 S(dded a discussion of unf)
-2.75 E(airness for Drop-T)-.11 E(ail with a queue)-.88 E(in bytes.)
111.8 527 Q(Changes from draft-ietf-dccp-tfrc-v)103.55 553 Q
(oip-00.txt:)-.22 E 2.75(*A)106.3 566 S(dded more simulations.)-2.75 E
2.75(*A)106.3 579 S(dded a Related W)-2.75 E(ork section.)-.88 E(Flo)72
733 Q(yd/K)-.11 E 333.671(ohler [P)-.385 F(age 3])-.165 E 0 Cg EP
%%Page: 4 4
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Times-Bold@0 SF -1.012
(Ta)72 85 S(ble of Contents)1.012 E/F2 10/Times-Roman@0 SF(1. Con)100.8
110 Q -.15(ve)-.4 G(ntions).15 E F0 -2.25(.............................\
.......................................................................\
................. ....)4.94 F F2(5)A(2. Introduction)100.8 122 Q F0
-2.25(.................................................................\
..................................................... ....)2.76 F F2(5)A
(3. TFRC-SP Congestion Control)100.8 134 Q F0 -2.25(...................\
.......................................................................\
.. ....)3.12 F F2(7)A(4. TFRC-SP Discussion)100.8 146 Q F0 -2.25(......\
.......................................................................\
........................... ....)4.85 F F2(8)A
(4.1. Response Functions and Throughput Equations)115.8 158 Q F0 -2.25
(........................................................... ....)2.51 F
F2(8)A(4.2. Accounting for Header Size)115.8 170 Q F0 -1.75(...........\
.......................................................................\
..... ..)3.04 F F2(12)A(4.3. The TFRC-SP Min Interv)115.8 182 Q(al)-.25
E F0 -1.75(............................................................\
........................... ..)4.92 F F2(12)A(4.4. Counting P)115.8 194
Q(ack)-.15 E(et Losses)-.1 E F0 -1.75(.................................\
............................................................ ..)2.59 F
F2(13)A(4.5. The Nominal P)115.8 206 Q(ack)-.15 E(et Size)-.1 E F0 -1.75
(......................................................................\
.................... ..)5.02 F F2(14)A(4.6. The Calculated Loss Interv)
115.8 218 Q(al Length for Short Loss Interv)-.25 E(als)-.25 E F0 -1.75
(....................................... ..)3.6 F F2(15)A
(5. A Comparison with RFC 3714)100.8 230 Q F0 -1.75(...................\
.......................................................................\
. ..)3.66 F F2(16)A(6. TFRC-SP with Applications that Modify the P)100.8
242 Q(ack)-.15 E(et Size)-.1 E F0 -1.75
(..................................................... ..)4.25 F F2(16)A
(7. Simulations)100.8 254 Q F0 -1.75(..................................\
.......................................................................\
............. ..)4.41 F F2(17)A(8. General Discussion)100.8 266 Q F0
-1.75(.................................................................\
.......................................... ..)4.41 F F2(17)A
(9. Security Considerations)100.8 278 Q F0 -1.75(......................\
.......................................................................\
........ ..)3.12 F F2(18)A(10. IAN)100.8 290 Q 2.5(AC)-.35 G
(onsiderations)-2.5 E F0 -1.75(........................................\
.............................................................. ..)4.06 F
F2(18)A(11. Conclusions)100.8 302 Q F0 -1.75(..........................\
.......................................................................\
................... ..)3.25 F F2(19)A(12. Thanks)100.8 314 Q F0 -1.75(.\
.......................................................................\
................................................... ..)4.01 F F2(19)A
(A. Appendix: Related W)100.8 326 Q(ork on Small-P)-.8 E(ack)-.15 E
(et V)-.1 E(ariants of TFRC)-1.11 E F0 -1.75
(............................................... ..)3.8 F F2(19)A
(B. Simulation Results)100.8 338 Q F0 -1.75(...........................\
.......................................................................\
......... ..)4.93 F F2(20)A(B.1. Simulations with Con\214gured P)115.8
350 Q(ack)-.15 E(et Drop Rates)-.1 E F0 -1.75
(......................................................... ..)3.53 F F2
(20)A(B.2. Simulations with Con\214gured Byte Drop Rates)115.8 362 Q F0
-1.75(............................................................ ..)
2.8 F F2(23)A(B.3. P)115.8 374 Q(ack)-.15 E(et Dropping Beha)-.1 E
(vior at Routers with Drop-T)-.2 E(ail Queues)-.8 E F0 -1.75
(...................................... ..)4.04 F F2(25)A(B.4. P)115.8
386 Q(ack)-.15 E(et Dropping Beha)-.1 E(vior at Routers with A)-.2 E(QM)
-.55 E F0 -1.75
(........................................................ ..)3.44 F F2
(29)A(C. Appendix: Exploring Possible Oscillations in the Loss Ev)100.8
398 Q(ent Rate)-.15 E F0 -1.75
(....................................... ..)4.04 F F2(33)A
(D. Appendix: A Discussion of P)100.8 410 Q(ack)-.15 E(et Size and P)-.1
E(ack)-.15 E(et Dropping)-.1 E F0 -1.75
(.............................................. ..)3.51 F F2(34)A
(Normati)100.8 422 Q .3 -.15(ve R)-.25 H(eferences).15 E F0 -1.75(.....\
.......................................................................\
............................... ..)3.16 F F2(34)A(Informati)100.8 434 Q
.3 -.15(ve R)-.25 H(eferences).15 E F0 -1.75(..........................\
.......................................................................\
........ ..)4.22 F F2(34)A(Authors' Addresses)100.8 446 Q F0 -1.75(....\
.......................................................................\
................................... ..)5.05 F F2(35)A(Full Cop)100.8 458
Q(yright Statement)-.1 E F0 -1.75(.....................................\
................................................................. ..)
4.62 F F2(35)A(Intellectual Property)100.8 470 Q F0 -1.75(.............\
.......................................................................\
......................... ..)5.02 F F2(36)A F0(Flo)72 733 Q(yd/K)-.11 E
333.671(ohler [P)-.385 F(age 4])-.165 E 0 Cg EP
%%Page: 5 5
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Times-Bold@0 SF(1.)72 85
Q/F2 14/Times-Bold@0 SF(Con)5.5 E -.14(ve)-.56 G(ntions).14 E F0(The k)
100.8 101.6 Q .33 -.165(ey w)-.11 H(ords "MUST", "MUST NO).055 E
(T", "REQ)-.44 E(UIRED", "SHALL", "SHALL NO)-.11 E(T",)-.44 E
("SHOULD", "SHOULD NO)100.8 114.6 Q(T", "RECOMMENDED", "MA)-.44 E
(Y", and "OPTION)-1.155 E(AL" in this)-.385 E
(document are to be interpreted as described in [RFC2119].)100.8 127.6 Q
F1(2.)72 153.6 Q F2(Intr)5.5 E(oduction)-.252 E F0
(This document speci\214es TFRC-SP)100.8 170.2 Q 2.75(,ae)-1.221 G
(xperimental, Small-P)-2.915 E(ack)-.165 E(et v)-.11 E
(ariant of TCP-friendly)-.275 E(rate control \(TFRC\) [RFC3448].)100.8
183.2 Q(TFRC w)100.8 199.8 Q(as designed to be reasonably f)-.11 E
(air when competing for bandwidth with TCP \215o)-.11 E(ws,)-.275 E -.22
(bu)100.8 212.8 S 2.75(tt).22 G 2.75(oa)-2.75 G -.22(vo)-2.97 G
(id the abrupt changes in the sending rate characteristic of TCP').22 E
2.75(sc)-.605 G(ongestion)-2.75 E(control mechanisms.)100.8 225.8 Q
(TFRC is intended for applications such as streaming media)5.5 E
(applications where a relati)100.8 238.8 Q -.165(ve)-.275 G
(ly smooth sending rate is of importance.).165 E(Con)5.5 E -.165(ve)-.44
G(ntional TFRC).165 E(measures loss rates by estimating the loss e)100.8
251.8 Q -.165(ve)-.275 G(nt ratio as described in [RFC3448], and uses)
.165 E(this loss e)100.8 264.8 Q -.165(ve)-.275 G
(nt rate to determine the sending rate in pack).165 E
(ets per round-trip time.)-.11 E(This has)5.5 E
(consequences for the rate that a TFRC \215o)100.8 277.8 Q 2.75(wc)-.275
G(an achie)-2.75 E .33 -.165(ve w)-.275 H(hen sharing a bottleneck with)
.165 E(lar)100.8 290.8 Q(ge-pack)-.198 E(et TCP \215o)-.11 E 2.75
(ws. In)-.275 F(particular)2.75 E 2.75(,al)-.44 G -.275(ow)-2.75 G
(-bandwidth, small-pack).275 E(et TFRC \215o)-.11 E 2.75(ws)-.275 G
(haring a)-2.75 E(bottleneck with high-bandwidth, lar)100.8 303.8 Q
(ge-pack)-.198 E(et TCP \215o)-.11 E(ws may be forced to slo)-.275 E
2.75(wd)-.275 G -.275(ow)-2.75 G(n, e).275 E -.165(ve)-.275 G(n).165 E
(though the TFRC application')100.8 316.8 Q 2.75(sn)-.605 G
(ominal rate in bytes per second is less than the rate)-2.75 E(achie)
100.8 329.8 Q -.165(ve)-.275 G 2.75(db).165 G 2.75(yt)-2.75 G
(he TCP \215o)-2.75 E 2.75(ws. Intuiti)-.275 F -.165(ve)-.275 G(ly).165
E 2.75(,t)-.715 G(his w)-2.75 E(ould be "f)-.11 E(air" only if the netw)
-.11 E(ork limitation)-.11 E -.11(wa)100.8 342.8 S 2.75(si).11 G 2.75
(np)-2.75 G(ack)-2.75 E(ets per second \(such as a routing lookup\), ra\
ther than bytes per second \(such as)-.11 E(link bandwidth\).)100.8
355.8 Q(Con)5.5 E -.165(ve)-.44 G(ntional wisdom is that man).165 E 2.75
(yo)-.165 G 2.75(ft)-2.75 G(he netw)-2.75 E(ork limitations in today')
-.11 E(s)-.605 E(Internet are in bytes per second, e)100.8 368.8 Q -.165
(ve)-.275 G 2.75(nt).165 G(hough the netw)-2.75 E
(ork limitations of the future might)-.11 E(mo)100.8 381.8 Q .33 -.165
(ve b)-.165 H(ack to).165 E -.11(wa)-.275 G(rds limitations in pack).11
E(ets per second.)-.11 E(TFRC-SP is intended for \215o)100.8 398.4 Q
(ws that need to send frequent small pack)-.275 E(ets, with less than)
-.11 E(1500 bytes per pack)100.8 411.4 Q
(et, limited by a minimum interv)-.11 E(al between pack)-.275 E
(ets of 10 ms.)-.11 E(It will)5.5 E
(better support applications that do not w)100.8 424.4 Q
(ant their sending rates in bytes per second to suf)-.11 E(fer)-.275 E
(from their use of small pack)100.8 437.4 Q 2.75(ets. This)-.11 F -.275
(va)2.75 G(riant is restricted to applications that send pack).275 E
(ets)-.11 E(no more than once e)100.8 450.4 Q -.165(ve)-.275 G
(ry 10 ms \(the Min Interv).165 E(al or minimum interv)-.275 E 2.75
(al\). Gi)-.275 F -.165(ve)-.275 G 2.75(nt).165 G(his)-2.75 E
(restriction, TFRC-SP ef)100.8 463.4 Q(fecti)-.275 E -.165(ve)-.275 G
(ly calculates the TFRC f).165 E(air rate as if the bottleneck)-.11 E
(restriction w)100.8 476.4 Q(as in bytes per second.)-.11 E
(Applications using TFRC-SP could ha)5.5 E .33 -.165(ve a \214)-.22 H
-.165(xe).165 G 2.75(do).165 G(r)-2.75 E(naturally-v)100.8 489.4 Q
(arying pack)-.275 E(et size, or could v)-.11 E(ary their pack)-.275 E
(et size in response to congestion.)-.11 E
(Applications that are not willing to be limited by a minimum interv)
100.8 502.4 Q(al of 10 ms. between)-.275 E(pack)100.8 515.4 Q
(ets, or that w)-.11 E(ant to send pack)-.11 E(ets lar)-.11 E
(ger than 1500 bytes, should not use TFRC-SP)-.198 E(.)-1.221 E(Ho)100.8
528.4 Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, f).165 H
(or applications with a minimum interv).44 E
(al of at least 10 ms. between pack)-.275 E(ets and)-.11 E
(with data pack)100.8 541.4 Q(ets of at most 1500 bytes, the performanc\
e of TFRC-SP should be at least as)-.11 E(good as that from TFRC.)100.8
554.4 Q(RFC 3448, the protocol speci\214cation for TFRC, stated that TF\
RC-PS \(for TFRC-)100.8 571 Q -.165(Pa)100.8 584 S(ck).165 E
(etSize\), a v)-.11 E(ariant of TFRC for applications that ha)-.275 E
.33 -.165(ve a \214)-.22 H -.165(xe).165 G 2.75(ds).165 G(ending rate b)
-2.75 E(ut v)-.22 E(ary their)-.275 E(pack)100.8 597 Q
(et size in response to congestion, w)-.11 E
(ould be speci\214ed in a later document.)-.11 E(This)5.5 E
(document instead speci\214es TFRC-SP)100.8 610 Q 2.75(,av)-1.221 G
(ariant of TFRC designed for applications that)-3.025 E(send small pack)
100.8 623 Q(ets, where applications could either ha)-.11 E .33 -.165
(ve a \214)-.22 H -.165(xe).165 G 2.75(do).165 G 2.75(rv)-2.75 G
(arying pack)-3.025 E(et size or)-.11 E(could adapt their pack)100.8 636
Q(et size in response to congestion.)-.11 E(Ho)5.5 E(we)-.275 E -.165
(ve)-.275 G .88 -.44(r, a).165 H 2.75(sd).44 G(iscussed in Section 6)
-2.75 E(of this document, there are man)100.8 649 Q 2.75(yq)-.165 G
(uestions about ho)-2.75 E 2.75(ws)-.275 G(uch an adapti)-2.75 E .33
-.165(ve a)-.275 H(pplication w).165 E(ould)-.11 E
(use TFRC-SP that are be)100.8 662 Q
(yond the scope of this document, and that w)-.165 E(ould need to be)
-.11 E(addressed in documents that are more application-speci\214c.)
100.8 675 Q(Flo)72 733 Q(yd/K)-.11 E 284.171(ohler Section)-.385 F 2.75
(2. [P)2.75 F(age 5])-.165 E 0 Cg EP
%%Page: 6 6
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(TFRC-SP is moti)100.8 85 Q
-.275(va)-.275 G(ted in part by the approach in RFC 3714, which ar).275
E(gues that it is)-.198 E(acceptable for V)100.8 98 Q(oIP \215o)-1.419 E
(ws to assume that the netw)-.275 E
(ork limitation is in bytes per second)-.11 E(\(Bps\) rather in pack)
100.8 111 Q(ets per second \(pps\), and to ha)-.11 E .33 -.165(ve t)-.22
H(he same sending rate in bytes per).165 E(second as a TCP \215o)100.8
124 Q 2.75(ww)-.275 G(ith 1500-byte pack)-2.75 E(ets and the same pack)
-.11 E(et drop rate.)-.11 E(RFC 3714)5.5 E(states the follo)100.8 137 Q
(wing:)-.275 E("While the ideal w)129.6 153.6 Q(ould be to ha)-.11 E .33
-.165(ve a t)-.22 H(ransport protocol that is able to detect whether)
.165 E(the bottleneck links along the path are limited in Bps or in pps\
, and to respond)129.6 166.6 Q(appropriately when the limitation is in \
pps, such an ideal is hard to achie)129.6 179.6 Q -.165(ve)-.275 G 2.75
(.W).165 G(e)-3.63 E -.11(wo)129.6 192.6 S(uld not w).11 E
(ant to delay the deplo)-.11 E(yment of congestion control for telephon)
-.11 E 2.75(yt)-.165 G(raf)-2.75 E<8c63>-.275 E
(until such an ideal could be accomplished.)129.6 205.6 Q
(In addition, we note that the current TCP)5.5 E
(congestion control mechanisms are themselv)129.6 218.6 Q(es not v)-.165
E(ery ef)-.165 E(fecti)-.275 E .33 -.165(ve i)-.275 H 2.75(na).165 G
2.75(ne)-2.75 G -.44(nv)-2.75 G(ironment).44 E
(where there is a limitation along the re)129.6 231.6 Q -.165(ve)-.275 G
(rse path in pps.).165 E(While the TCP)5.5 E(mechanisms do pro)129.6
244.6 Q(vide an incenti)-.165 E .33 -.165(ve t)-.275 H 2.75(ou).165 G
(se lar)-2.75 E(ge data pack)-.198 E(ets, TCP does not include)-.11 E
(an)129.6 257.6 Q 2.75(ye)-.165 G -.275(ff)-2.75 G(ecti).275 E .33 -.165
(ve c)-.275 H(ongestion control mechanisms for the stream of small).165
E(ackno)129.6 270.6 Q(wledgement pack)-.275 E(ets on the re)-.11 E -.165
(ve)-.275 G(rse path.).165 E(Gi)5.5 E -.165(ve)-.275 G 2.75(nt).165 G
(he ar)-2.75 E(guments abo)-.198 E -.165(ve)-.165 G 2.75(,i).165 G 2.75
(ts)-2.75 G(eems)-2.75 E(acceptable to us to assume a netw)129.6 283.6 Q
(ork limitation in Bps rather than in pps in)-.11 E
(considering the minimum sending rate of telephon)129.6 296.6 Q 2.75(yt)
-.165 G(raf)-2.75 E(\214c.")-.275 E -.385(Tr)100.8 313.2 S(anslating th\
e discussion in [RFC3714] to the congestion control mechanisms of TFRC,\
 it).385 E(seems acceptable to standardize a v)100.8 326.2 Q
(ariant of TFRC that allo)-.275 E(ws lo)-.275 E(w-bandwidth \215o)-.275
E(ws)-.275 E(sending small pack)100.8 339.2 Q(ets to achie)-.11 E .33
-.165(ve a r)-.275 H(ough f).165 E(airness with TCP \215o)-.11 E
(ws in terms of the sending)-.275 E
(rate in Bps, rather than in terms of the sending rate in pps.)100.8
352.2 Q(This is accomplished by TFRC-)5.5 E(SP)100.8 365.2 Q 2.75(,as)
-1.221 G(mall modi\214cation to TFRC, as described belo)-2.75 E -.715
(w.)-.275 G(Maintaining incenti)100.8 381.8 Q -.165(ve)-.275 G 2.75(sf)
.165 G(or lar)-2.75 E(ge pack)-.198 E
(ets: Because the bottlenecks in the netw)-.11 E(ork in f)-.11 E
(act can)-.11 E(include limitations in pps as well as in Bps, we pay sp\
ecial attention to the potential)100.8 394.8 Q
(dangers of encouraging a lar)100.8 407.8 Q(ge deplo)-.198 E
(yment of best-ef)-.11 E(fort traf)-.275 E
(\214c in the Internet consisting)-.275 E(entirely of small pack)100.8
420.8 Q 2.75(ets. This)-.11 F
(is discussed in more detail in Section 4.3. In addition, as)2.75 E(ag)
100.8 433.8 Q(ain discussed in Section 4.3, TFRC-SP includes the limita\
tion of the Min Interv)-.055 E(al)-.275 E(between pack)100.8 446.8 Q
(ets of 10 ms.)-.11 E -.165(Pa)100.8 463.4 S(ck).165 E
(et drop rates as a function of pack)-.11 E
(et size: TFRC-SP essentially assumes that the small-)-.11 E(pack)100.8
476.4 Q(et TFRC-SP \215o)-.11 E 2.75(wr)-.275 G(ecei)-2.75 E -.165(ve)
-.275 G 2.75(sr).165 G(oughly the same pack)-2.75 E
(et drop rate as a lar)-.11 E(ge-pack)-.198 E(et TFRC)-.11 E
(or TCP \215o)100.8 489.4 Q 4.18 -.715(w. A)-.275 H 2.75(sw).715 G 2.75
(es)-2.75 G(ho)-2.75 E 1.43 -.715(w, t)-.275 H
(his assumption is not necessarily correct for all en).715 E(vironments)
-.44 E(in the Internet.)100.8 502.4 Q
(Initializing the Loss History after the First Loss Ev)100.8 519 Q
(ent: Section 6.3.1 of RFC 3448 speci\214es)-.165 E(that the TFRC recei)
100.8 532 Q -.165(ve)-.275 G 2.75(ri).165 G
(nitializes the loss history after the \214rst loss e)-2.75 E -.165(ve)
-.275 G(nt by calculating the).165 E(loss interv)100.8 545 Q(al that w)
-.275 E(ould be required to produce the recei)-.11 E .33 -.165(ve r)
-.275 H(ate measured o).165 E -.165(ve)-.165 G 2.75(rt).165 G(he most)
-2.75 E(recent round-trip time.)100.8 558 Q
(In calculating this loss interv)5.5 E(al, TFRC-SP uses the se)-.275 E
(gment size of)-.165 E(1460 bytes, rather than the actual se)100.8 571 Q
(gment size used in the connection.)-.165 E(Calculating the loss e)100.8
587.6 Q -.165(ve)-.275 G
(nt rate for TFRC-SP: TFRC-SP requires a modi\214cation in TFRC').165 E
(s)-.605 E(calculation of the loss e)100.8 600.6 Q -.165(ve)-.275 G
(nt rate, because a TFRC-SP connection can send man).165 E 2.75(ys)-.165
G(mall)-2.75 E(pack)100.8 613.6 Q
(ets when a standard TFRC or TCP connection w)-.11 E
(ould send a single lar)-.11 E(ge pack)-.198 E 2.75(et. It)-.11 F(is)
2.75 E(not possible for a standard TFRC or TCP connection to repeatedly\
 send multiple pack)100.8 626.6 Q(ets)-.11 E
(per round-trip time in the f)100.8 639.6 Q(ace of a high pack)-.11 E
(et drop rate.)-.11 E(As a result, TCP and standard)5.5 E
(TFRC only respond to a single loss e)100.8 652.6 Q -.165(ve)-.275 G
(nt per round-trip time, and are still able to detect).165 E
(and respond to increasingly hea)100.8 665.6 Q(vy pack)-.22 E
(et loss rates.)-.11 E(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44
(r, i).165 H 2.75(nah).44 G(ighly-congested)-2.75 E(en)100.8 678.6 Q
(vironment, when a TCP connection might be sending, on a)-.44 E -.165
(ve)-.22 G(rage, one lar).165 E(ge pack)-.198 E(et per)-.11 E(round-tri\
p time, a corresponding TFRC-SP connection might be sending man)100.8
691.6 Q 2.75(ys)-.165 G(mall)-2.75 E(Flo)72 733 Q(yd/K)-.11 E 284.171
(ohler Section)-.385 F 2.75(2. [P)2.75 F(age 6])-.165 E 0 Cg EP
%%Page: 7 7
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(pack)100.8 85 Q
(ets per round-trip time.)-.11 E(As a result, in order to maintain f)5.5
E(airness with TCP)-.11 E 2.75(,a)-1.221 G(nd to be)-2.75 E
(able to detect changes in the de)100.8 98 Q
(gree of congestion, TFRC-SP needs to be sensiti)-.165 E .33 -.165(ve t)
-.275 H 2.75(ot).165 G(he)-2.75 E(actual pack)100.8 111 Q
(et drop rate during periods of high congestion.)-.11 E
(This is discussed in more detail)5.5 E(in the section belo)100.8 124 Q
-.715(w.)-.275 G/F1 11/Times-Bold@0 SF(3.)72 150 Q/F2 14/Times-Bold@0 SF
(TFRC-SP Congestion Contr)5.5 E(ol)-.252 E F0
(TFRC uses the TCP throughput equation gi)100.8 166.6 Q -.165(ve)-.275 G
2.75(ni).165 G 2.75(nS)-2.75 G(ection 3.1 of [RFC3448], which gi)-2.75 E
-.165(ve)-.275 G(s).165 E(the allo)100.8 179.6 Q
(wed sending rate X in bytes per second as a function of the loss e)
-.275 E -.165(ve)-.275 G(nt rate, se).165 E(gment)-.165 E
(size, and round-trip time.)100.8 192.6 Q
([RFC3448] speci\214es that the se)5.5 E(gment size s used in the)-.165
E(throughput equation should be the se)100.8 205.6 Q
(gment size used by the application, or the estimated)-.165 E(mean se)
100.8 218.6 Q(gment size if there are v)-.165 E(ariations in the se)
-.275 E(gment size depending on the data.)-.165 E(This)5.5 E(gi)100.8
231.6 Q -.165(ve)-.275 G 2.75(sr).165 G(ough f)-2.75 E
(airness with TCP \215o)-.11 E(ws using the same se)-.275 E(gment size.)
-.165 E(TFRC-SP changes this beha)100.8 248.2 Q(vior in the follo)-.22 E
(wing w)-.275 E(ays.)-.11 E 12.65<8354>100.8 264.8 S(he nominal se)
-12.65 E(gment size: The nominal se)-.165 E
(gment size s is set to 1460 bytes.)-.165 E -.165(Fo)5.5 G(llo).165 E
(wing)-.275 E([RFC3714], this pro)117.3 277.8 Q(vides a goal of f)-.165
E(airness, in terms of the sending rate in bytes per)-.11 E
(second, with a TCP \215o)117.3 290.8 Q 2.75(ww)-.275 G
(ith 1460 bytes of application data per pack)-2.75 E(et b)-.11 E
(ut with the)-.22 E(same pack)117.3 303.8 Q(et drop rate.)-.11 E
(Instead of using a nominal se)117.3 320.4 Q
(gment size of 1460 bytes, an alternate possibility w)-.165 E(ould)-.11
E(ha)117.3 333.4 Q .33 -.165(ve b)-.22 H
(een for TFRC-SP to determine the actual Maximum Se).165 E
(gment Size \(MSS\) of the)-.165 E
(path, and to use this for the nominal se)117.3 346.4 Q(gment size.)
-.165 E(While most paths ha)5.5 E .33 -.165(ve a)-.22 H 2.75(nM).165 G
(SS of)-2.75 E(1460 bytes, some paths ha)117.3 359.4 Q .33 -.165(ve a s)
-.22 H(lightly smaller MSS due to tunnels, IPv6, and the lik).165 E(e,)
-.11 E(and some other paths ha)117.3 372.4 Q .33 -.165(ve a s)-.22 H
(igni\214cantly smaller MSS of only 536 bytes.).165 E(Due to the)5.5 E
(complications of estimating the MSS of the path, and to the f)117.3
385.4 Q(act that most paths support)-.11 E
(an MSS of at least 536 bytes, we ha)117.3 398.4 Q .33 -.165(ve d)-.22 H
(ecided to use a nominal se).165 E(gment size of 1460)-.165 E
(bytes for TFRC-SP)117.3 411.4 Q(.)-1.221 E 12.65<8354>100.8 428 S
(aking pack)-13.53 E(et headers into account: The allo)-.11 E
(wed transmit rate X in bytes per second is)-.275 E(reduced by a f)117.3
441 Q(actor that accounts for pack)-.11 E(et header size.)-.11 E
(This gi)5.5 E -.165(ve)-.275 G 2.75(st).165 G(he application)-2.75 E
(some incenti)117.3 454 Q -.165(ve)-.275 G 2.75(,b).165 G -.165(ey)-2.75
G(ond the Min Interv).165 E(al, not to use unnecessarily small pack)
-.275 E 2.75(ets. In)-.11 F(particular)117.3 467 Q 2.75(,w)-.44 G 2.75
(ei)-2.75 G(ntroduce a ne)-2.75 E 2.75(wp)-.275 G
(arameter H, which represents the e)-2.75 E(xpected size in bytes)-.165
E(of netw)117.3 480 Q
(ork and transport headers to be used on the TFRC connection')-.11 E
2.75(sp)-.605 G(ack)-2.75 E 2.75(ets. This)-.11 F(is)2.75 E
(used to reduce the allo)117.3 493 Q(wed transmit rate X as follo)-.275
E(ws:)-.275 E 2.75(X:)117.3 509.6 S 2.75(=X*s)-2.75 G
(_true / \(s_true + H\),)-2.75 E(where s_true is the true a)117.3 526.2
Q -.165(ve)-.22 G(rage data se).165 E
(gment size for the connection in bytes,)-.165 E -.165(ex)117.3 539.2 S
(cluding the transport and netw).165 E(ork headers.)-.11 E
(Section 4.1 of RFC 3448 states that)5.5 E(where the pack)117.3 552.2 Q
(et size v)-.11 E
(aries naturally with the data, an estimate of the mean se)-.275 E
(gment)-.165 E(size can be used for s_true.)117.3 565.2 Q
(As suggested in Section 4.1 of [RFC3448bis], when an)5.5 E
(estimate of the mean se)117.3 578.2 Q
(gment size is used for s_true, the estimate SHOULD be)-.165 E
(calculated o)117.3 591.2 Q -.165(ve)-.165 G 2.75(ra).165 G 2.75(tl)
-2.75 G(east the last four loss interv)-2.75 E 2.75(als. Ho)-.275 F(we)
-.275 E -.165(ve)-.275 G .88 -.44(r, t).165 H(his document does not).44
E(specify a speci\214c algorithm for estimating the mean se)117.3 604.2
Q(gment size.)-.165 E(The H parameter is set to the constant 40 bytes.)
117.3 620.8 Q(Thus, if the TFRC-SP application used)5.5 E
(40-byte data se)117.3 633.8 Q(gments, the allo)-.165 E
(wed transmit rate X w)-.275 E(ould be halv)-.11 E
(ed to account for the)-.165 E -.11(fa)117.3 646.8 S
(ct that half of the sending rate w).11 E(ould be used by the headers.)
-.11 E(Section 4.2 justi\214es this)5.5 E 2.75(de\214nition. Ho)117.3
659.8 R(we)-.275 E -.165(ve)-.275 G .88 -.44(r, a c).165 H
(onnection using TFRC-SP MA).44 E 2.75(Yi)-1.155 G
(nstead use a more precise)-2.75 E
(estimate of H, based on the actual netw)117.3 672.8 Q
(ork and transport headers to be used on the)-.11 E(connection')117.3
685.8 Q 2.75(sp)-.605 G(ack)-2.75 E 2.75(ets. F)-.11 F(or e)-.165 E
(xample, a DCCP connection [DCCP] o)-.165 E -.165(ve)-.165 G 2.75(rI)
.165 G(Pv4, where data)-2.75 E(Flo)72 733 Q(yd/K)-.11 E 284.171
(ohler Section)-.385 F 2.75(3. [P)2.75 F(age 7])-.165 E 0 Cg EP
%%Page: 8 8
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(pack)117.3 85 Q
(ets use the DCCP-Data pack)-.11 E
(et type, and there are no IP or DCCP options, could)-.11 E
(set H to 20 + 12 = 32 bytes.)117.3 98 Q(If the TFRC implementation kno)
5.5 E(ws that the IP layer is)-.275 E
(using IPv6 instead of IPv4, then the connection using TFRC-SP MA)117.3
111 Q 2.75(Yu)-1.155 G(se an estimate)-2.75 E(of 40 bytes instead of 60\
 bytes for H, for simplicity of implementation.)117.3 124 Q 12.65<834d>
100.8 140.6 S(easuring the loss e)-12.65 E -.165(ve)-.275 G
(nt rate in times of high loss: During short loss interv).165 E
(als \(those at)-.275 E(most tw)117.3 153.6 Q 2.75(or)-.11 G(ound-trip \
times\), the loss rate is computed by counting the actual number of)
-2.75 E(pack)117.3 166.6 Q(ets lost or mark)-.11 E
(ed, not by counting at most one loss e)-.11 E -.165(ve)-.275 G
(nt per loss interv).165 E(al.)-.275 E -.44(Wi)117.3 179.6 S
(thout this change, TFRC-SP could send multiple pack).44 E
(ets per round-trip time e)-.11 E -.165(ve)-.275 G 2.75(ni).165 G(n)
-2.75 E(the f)117.3 192.6 Q(ace of hea)-.11 E
(vy congestion, for a steady-state beha)-.22 E(vior of multiple pack)
-.22 E(ets dropped)-.11 E(each round-trip time.)117.3 205.6 Q
(In standard TFRC, the TFRC recei)117.3 222.2 Q -.165(ve)-.275 G 2.75
(re).165 G(stimates the loss e)-2.75 E -.165(ve)-.275 G
(nt rate by calculating the).165 E -2.475 -.22(av e)117.3 235.2 T
(rage loss interv).22 E(al in pack)-.275 E(ets, and in)-.11 E -.165(ve)
-.44 G(rting to get the loss e).165 E -.165(ve)-.275 G(nt rate.).165 E
(Thus, for a)5.5 E(short loss interv)117.3 248.2 Q(al with N pack)-.275
E(ets and K losses, standard TFRC calculates the size of)-.11 E
(that loss interv)117.3 261.2 Q(al as N pack)-.275 E(ets, contrib)-.11 E
(uting to a loss e)-.22 E -.165(ve)-.275 G(nt rate of 1/N.).165 E(Ho)5.5
E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, f).165 H(or).44 E(TFRC-SP)
117.3 274.2 Q 2.75(,f)-1.221 G(or small loss interv)-2.75 E
(als of at most tw)-.275 E 2.75(or)-.11 G
(ound-trip times, if the loss interv)-2.75 E(al)-.275 E
(consists of N pack)117.3 287.2 Q
(ets including K losses, the size of the loss interv)-.11 E
(al is calculated as)-.275 E(N/K, contrib)117.3 300.2 Q
(uting to a loss e)-.22 E -.165(ve)-.275 G
(nt rate of K/N instead of 1/N.).165 E
(Section 5.4 of RFC 3448 speci\214es that the calculation of the a)117.3
316.8 Q -.165(ve)-.22 G(rage loss interv).165 E(al)-.275 E
(includes the most recent loss interv)117.3 329.8 Q
(al only if this increases the calculated a)-.275 E -.165(ve)-.22 G
(rage loss).165 E(interv)117.3 342.8 Q 2.75(al. Ho)-.275 F(we)-.275 E
-.165(ve)-.275 G .88 -.44(r, f).165 H(or TFRC-SP).44 E 2.75(,i)-1.221 G
(ncluding the most recent loss interv)-2.75 E(al can increase)-.275 E
(the calculated a)117.3 355.8 Q -.165(ve)-.22 G(rage loss interv).165 E
(al too much if the most recent loss interv)-.275 E(al ends up)-.275 E
(being a short loss interv)117.3 368.8 Q(al with multiple pack)-.275 E
(et drops.)-.11 E(Therefore, TFRC-SP adds the)5.5 E
(restriction that the calculation of the a)117.3 381.8 Q -.165(ve)-.22 G
(rage loss interv).165 E(al can include the most recent)-.275 E
(loss interv)117.3 394.8 Q(al only if more than tw)-.275 E 2.75(or)-.11
G(ound-trip times ha)-2.75 E .33 -.165(ve p)-.22 H(assed since the be)
.165 E(ginning of)-.165 E(that loss interv)117.3 407.8 Q(al.)-.275 E
12.65<8341>100.8 424.4 S(minimum interv)-9.9 E(al between pack)-.275 E
(ets: TFRC-SP enforces a Min Interv)-.11 E(al between)-.275 E(pack)117.3
437.4 Q(ets of 10 ms.)-.11 E 2.75<418d>5.5 G .55 -.275(ow t)-2.75 H
(hat wishes its transport protocol to e).275 E(xceed this Min Interv)
-.165 E(al)-.275 E(MUST use the con)117.3 450.4 Q -.165(ve)-.44 G
(ntional TFRC equations, rather than TFRC-SP).165 E 5.5(.T)-1.221 G
(he moti)-5.5 E -.275(va)-.275 G(tion).275 E(for this is discussed belo)
117.3 463.4 Q -.715(w.)-.275 G/F1 11/Times-Bold@0 SF(4.)72 489.4 Q/F2 14
/Times-Bold@0 SF(TFRC-SP Discussion)5.5 E F1(4.1.)72 515.4 Q/F3 13
/Times-Bold@0 SF(Response Functions and Thr)5.5 E(oughput Equations)
-.234 E F0(TFRC uses the TCP throughput equation gi)100.8 532 Q -.165
(ve)-.275 G 2.75(ni).165 G 2.75(n[)-2.75 G
(RFC3448], with the sending rate X in)-2.75 E(bytes per second as follo)
100.8 545 Q(ws:)-.275 E(s)208.3 564 Q 2.75(X=-)136.8 577 S
(------------------------------------------------------ ,)-2.75 E
(R*sqrt\(2*p/3\) + \(4*R* \(3*sqrt\(3*p/8\) * p * \(1+32*p^2\)\)\))147.8
590 Q(where:)100.8 609 Q/F4 11/Times-Italic@0 SF(s)129.6 625.6 Q F0
(is the pack)2.75 E(et size in bytes;)-.11 E F4(R)129.6 642.2 Q F0
(is the round trip time in seconds;)2.75 E F4(p)129.6 658.8 Q F0
(is the loss e)2.75 E -.165(ve)-.275 G
(nt rate, between 0 and 1.0, of the number of loss e).165 E -.165(ve)
-.275 G(nts as a fraction).165 E(of the number of pack)129.6 671.8 Q
(ets transmitted.)-.11 E(Flo)72 733 Q(yd/K)-.11 E 275.921(ohler Section)
-.385 F 2.75(4.1. [P)2.75 F(age 8])-.165 E 0 Cg EP
%%Page: 9 9
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(This equation uses an R)100.8
85 Q .396 -.198(TO o)-.66 H 2.75(f$).198 G
(4*R$, and assumes that the TCP connection sends an)-2.75 E(ackno)100.8
98 Q(wledgement for e)-.275 E -.165(ve)-.275 G(ry data pack).165 E(et.)
-.11 E(This equation essentially gi)100.8 114.6 Q -.165(ve)-.275 G 2.75
(st).165 G(he response function for TCP as well as for standard TFRC)
-2.75 E(\(modulo TCP')100.8 127.6 Q 2.75(sr)-.605 G
(ange of sender algorithms for setting the R)-2.75 E -.198(TO)-.66 G
2.75(\). As).198 F(sho)2.75 E(wn in T)-.275 E(able 1 of)-.88 E
([RFC3714], for high pack)100.8 140.6 Q
(et drop rates, this throughput equation gi)-.11 E -.165(ve)-.275 G 2.75
(sr).165 G(ough f)-2.75 E(airness with)-.11 E(the most aggressi)100.8
153.6 Q .33 -.165(ve p)-.275 H(ossible current TCP: a SA).165 E
(CK TCP \215o)-.44 E 2.75(wu)-.275 G(sing timestamps and ECN.)-2.75 E(B\
ecause it is not recommended for routers to use ECN-marking in highly-c\
ongested)100.8 166.6 Q(en)100.8 179.6 Q(vironments with high pack)-.44 E
(et dropping/marking rates [RFC3168] \(Section 7\), we note)-.11 E
(that it w)100.8 192.6 Q(ould be useful to ha)-.11 E .33 -.165(ve a t)
-.22 H(hroughput equation with a some).165 E(what more moderate)-.275 E
(sending rate for pack)100.8 205.6 Q(et drop rates of 40% and abo)-.11 E
-.165(ve)-.165 G(.).165 E(The ef)100.8 222.2 Q(fecti)-.275 E .33 -.165
(ve r)-.275 H(esponse function of TFRC-SP can be deri).165 E -.165(ve)
-.275 G 2.75(df).165 G(rom the TFRC response)-2.75 E
(function by using a se)100.8 235.2 Q
(gment size s of 1460 bytes, and using the loss e)-.165 E -.165(ve)-.275
G(nt rate actually).165 E -.165(ex)100.8 248.2 S
(perienced by the TFRC-SP \215o).165 E 4.18 -.715(w. I)-.275 H 2.75(na)
.715 G(ddition, for loss interv)-2.75 E(als of at most tw)-.275 E 2.75
(or)-.11 G(ound-trip)-2.75 E(times, the loss e)100.8 261.2 Q -.165(ve)
-.275 G
(nt rate for TFRC-SP is estimated by counting the actual number of lost)
.165 E(or mark)100.8 274.2 Q(ed pack)-.11 E
(ets, rather than by counting loss e)-.11 E -.165(ve)-.275 G 2.75
(nts. In).165 F(addition, the sending rate for)2.75 E
(TFRC-SP is constrained to be at most 100 pack)100.8 287.2 Q
(ets per second.)-.11 E -.165(Fo)100.8 303.8 S 2.75(ra).165 G 2.75(ne)
-2.75 G -.44(nv)-2.75 G(ironment with a \214x).44 E(ed pack)-.165 E
(et drop rate p, re)-.11 E -.055(ga)-.165 G(rdless of pack).055 E
(et size, the response)-.11 E(functions of TCP)100.8 316.8 Q 2.75(,T)
-1.221 G(FRC, and TFRC-SP are illustrated as follo)-2.75 E(ws, gi)-.275
E -.165(ve)-.275 G 2.75(ni).165 G 2.75(nK)-2.75 G(Bps \(Kilo)-2.75 E
(Bytes per second\), for a \215o)100.8 329.8 Q 2.75(ww)-.275 G
(ith a round-trip time of 100 ms:)-2.75 E/F1 11/Courier@0 SF 6.6
(<-- TCP)246 348.8 R(and Standard TFRC)6.6 E(-->)13.2 E 26.4
(Packet 14-byte)173.4 361.8 R 13.2(536-byte 1460-byte)26.4 F 13.2
(DropRate Segments Segments Segments)173.4 374.8 R 13.2
(-------- -------- -------- --------)173.4 387.8 R 26.4(0.00001 364.25)
173.4 400.8 R 19.8(2232.00 5967.49)26.4 F 26.4(0.00003 210.26)173.4
413.8 R 19.8(1288.41 3444.71)26.4 F 26.4(0.00010 115.09 705.25)173.4
426.8 R(1885.56)26.4 E 33(0.00030 66.33)173.4 439.8 R 19.8
(406.44 1086.67)33 F 33(0.00100 36.10)173.4 452.8 R 26.4(221.23 591.48)
33 F 33(0.00300 20.48)173.4 465.8 R 26.4(125.49 335.51)33 F 33
(0.01000 10.57 64.75)173.4 478.8 R(173.10)33 E 39.6(0.03000 5.21)173.4
491.8 R 33(31.90 85.28)39.6 F 39.6(0.10000 1.67)173.4 504.8 R 33
(10.21 27.28)39.6 F 39.6(0.20000 0.50 3.09 8.27)173.4 517.8 R 39.6
(0.30000 0.18 1.12 3.00)173.4 530.8 R 39.6(0.40000 0.08 0.48 1.30)173.4
543.8 R 39.6(0.50000 0.04 0.24 0.64)173.4 556.8 R
(Table 1: Response Function for TCP and TFRC.)166.8 582.8 Q
(Sending Rate in KBps, as a Function of Packet Drop Rate.)127.2 595.8 Q
F0(Flo)72 733 Q(yd/K)-.11 E 275.921(ohler Section)-.385 F 2.75(4.1. [P)
2.75 F(age 9])-.165 E 0 Cg EP
%%Page: 10 10
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<---------- TFRC-SP -------->)246 85 R 26.4(Packet 14-byte)173.4 98 R
13.2(536-byte 1460-byte)26.4 F 13.2(DropRate Segments Segments Segments)
173.4 111 R 13.2(-------- -------- -------- --------)173.4 124 R 39.6
(0.00001 5.40)173.4 137 R 26.4(53.60 150.00)39.6 F 39.6(0.00003 5.40)
173.4 150 R 26.4(53.60 150.00)39.6 F 39.6(0.00010 5.40)173.4 163 R 26.4
(53.60 150.00)39.6 F 39.6(0.00030 5.40)173.4 176 R 26.4(53.60 150.00)
39.6 F 39.6(0.00100 5.40)173.4 189 R 26.4(53.60 150.00)39.6 F 39.6
(0.00300 5.40)173.4 202 R 26.4(53.60 150.00)39.6 F 39.6(0.01000 5.40)
173.4 215 R 26.4(53.60 150.00)39.6 F 39.6(0.03000 5.40)173.4 228 R 33
(53.60 83.07)39.6 F 39.6(0.10000 5.40)173.4 241 R 33(26.58 26.58)39.6 F
39.6(0.20000 5.40 8.06 8.06)173.4 254 R 39.6(0.30000 2.93 2.93 2.93)
173.4 267 R 39.6(0.40000 1.26 1.26 1.26)173.4 280 R 39.6
(0.50000 0.63 0.63 0.63)173.4 293 R
(Table 2: Response Function for TFRC-SP.)180 319 Q
(Sending Rate in KBps, as a Function of Packet Drop Rate.)127.2 332 Q
(Maximum Sending Rate of 100 Packets per Second.)153.6 345 Q F0
(The calculations for T)100.8 367.6 Q(ables 1 and 2 use the pack)-.88 E
(et loss rate for an approximation for the)-.11 E(loss e)100.8 380.6 Q
-.165(ve)-.275 G(nt rate p.).165 E
(Scripts and graphs for the tables are a)5.5 E -.275(va)-.22 G
(ilable from [V).275 E 2.75(OIPSIMS]. As)-.44 F(the)2.75 E(well-kno)
100.8 393.6 Q(wn TCP response function in T)-.275 E(able 1 sho)-.88 E
(ws, the sending rate for TCP and)-.275 E(standard TFRC v)100.8 406.6 Q
(aries linearly with se)-.275 E(gment size.)-.165 E
(The TFRC-SP response function sho)5.5 E(wn)-.275 E(in T)100.8 419.6 Q
(able 2 re\215ects the maximum sending rate of a hundred pack)-.88 E
(ets per second; when not)-.11 E
(limited by this maximum sending rate, the TFRC-SP \215o)100.8 432.6 Q
2.75(wr)-.275 G(ecei)-2.75 E -.165(ve)-.275 G 2.75(st).165 G
(he same sending rate in)-2.75 E(KBps as the TCP \215o)100.8 445.6 Q
2.75(ww)-.275 G(ith 1460-byte se)-2.75 E(gments, gi)-.165 E -.165(ve)
-.275 G 2.75(nt).165 G(he same pack)-2.75 E(et drop rate.)-.11 E
(Simulations sho)100.8 458.6 Q(wing the TCP)-.275 E 2.75(,s)-1.221 G
(tandard TFRC, and TFRC-SP sending rates in response to a)-2.75 E
(con\214gured pack)100.8 471.6 Q(et drop rate are gi)-.11 E -.165(ve)
-.275 G 2.75(ni).165 G 2.75(nT)-2.75 G
(ables 7, 8, and 9, and are consistent with the)-3.63 E
(response functions sho)100.8 484.6 Q(wn here.)-.275 E(Flo)72 733 Q
(yd/K)-.11 E 270.421(ohler Section)-.385 F 2.75(4.1. [P)2.75 F(age 10])
-.165 E 0 Cg EP
%%Page: 11 11
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<-- TCP)259.2 85 R(and Standard TFRC)6.6 E(-->)13.2 E 52.8
(Byte 14-byte)173.4 98 R 13.2(536-byte 1460-byte)26.4 F 26.4
(DropRate Segments)173.4 111 R 13.2(Segments Segments)19.8 F 26.4
(-------- --------)173.4 124 R 13.2(-------- --------)19.8 F 26.4
(0.0000001 375.70 929.61)173.4 137 R(1518.75)26.4 E 26.4
(0.0000003 216.87 536.17 874.50)173.4 150 R 26.4
(0.0000010 118.72 292.64 474.53)173.4 163 R 33(0.0000030 68.43)173.4 176
R 26.4(167.28 266.90)33 F 33(0.0000100 37.27 88.56)173.4 189 R(134.09)33
E 33(0.0000300 21.17 46.67 62.00)173.4 202 R 33
(0.0001000 10.98 19.20 16.01)173.4 215 R 39.6(0.0003000 5.50 4.95 1.64)
173.4 228 R 39.6(0.0010000 1.91 0.37 0.14)173.4 241 R 39.6
(0.0030000 0.31 0.05 0.07)173.4 254 R 39.6(0.0100000 0.02 0.02 0.06)
173.4 267 R 39.6(0.0300000 0.00 0.02 0.06)173.4 280 R
(Table 3: Response Function for TCP and TFRC.)173.4 306 Q
(Sending Rate in KBps, as a Function of Byte Drop Rate.)133.8 319 Q 6.6
(<---------- TFRC-SP -------->)259.2 344 R 52.8(Byte 14-byte)173.4 357 R
13.2(536-byte 1460-byte)26.4 F 26.4(DropRate Segments)173.4 370 R 13.2
(Segments Segments)19.8 F 26.4(-------- --------)173.4 383 R 13.2
(-------- --------)19.8 F 39.6(0.0000001 5.40)173.4 396 R 26.4
(53.60 150.00)39.6 F 39.6(0.0000003 5.40)173.4 409 R 26.4(53.60 150.00)
39.6 F 39.6(0.0000010 5.40)173.4 422 R 26.4(53.60 150.00)39.6 F 39.6
(0.0000030 5.40)173.4 435 R 26.4(53.60 150.00)39.6 F 39.6
(0.0000100 5.40)173.4 448 R 26.4(53.60 130.61)39.6 F 39.6
(0.0000300 5.40)173.4 461 R 33(53.60 60.39)39.6 F 39.6(0.0001000 5.40)
173.4 474 R 33(50.00 15.59)39.6 F 39.6(0.0003000 5.40)173.4 487 R 39.6
(12.89 1.60)39.6 F 39.6(0.0010000 5.40 0.95 0.14)173.4 500 R 39.6
(0.0030000 4.94 0.12 0.06)173.4 513 R 39.6(0.0100000 0.33 0.06 0.06)
173.4 526 R 39.6(0.0300000 0.08 0.06 0.06)173.4 539 R
(Table 4: Response Function for TFRC-SP.)186.6 565 Q
(Sending Rate in KBps, as a Function of Byte Drop Rate.)133.8 578 Q
(Maximum Sending Rate of 100 Packets per Second.)153.6 591 Q F0 -.165
(Fo)100.8 613.6 S 2.75(rT).165 G(ables 3 and 4, the pack)-3.63 E
(et drop rate is calculated as 1-\(1-b\)^N, for a byte drop rate of b,)
-.11 E(and a pack)100.8 626.6 Q(et size of N bytes.)-.11 E
(These tables use the pack)5.5 E(et loss rate as an approximation for)
-.11 E(the loss e)100.8 639.6 Q -.165(ve)-.275 G(nt rate p.).165 E
(The TCP response functions sho)5.5 E(wn in T)-.275 E(able 3 for \214x)
-.88 E(ed byte drop)-.165 E(rates are rather dif)100.8 652.6 Q
(ferent from the response functions sho)-.275 E(wn in T)-.275 E
(able 1 for \214x)-.88 E(ed pack)-.165 E(et drop)-.11 E
(rates; with higher byte drop rates, a TCP connection can ha)100.8 665.6
Q .33 -.165(ve a h)-.22 H(igher sending rate using).165 E
(*smaller* pack)100.8 678.6 Q 2.75(ets. T)-.11 F(able 4 also sho)-.88 E
(ws that with \214x)-.275 E(ed byte drop rates, the sending rate for)
-.165 E(TFRC-SP can be signi\214cantly higher than that for TCP or stan\
dard TFRC, re)100.8 691.6 Q -.055(ga)-.165 G(rdless of the).055 E(Flo)72
733 Q(yd/K)-.11 E 270.421(ohler Section)-.385 F 2.75(4.1. [P)2.75 F
(age 11])-.165 E 0 Cg EP
%%Page: 12 12
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(TCP se)100.8 85 Q(gment size.)
-.165 E(This is because in this en)5.5 E(vironment, with small pack)-.44
E(ets, TFRC-SP)-.11 E(recei)100.8 98 Q -.165(ve)-.275 G 2.75(sas).165 G
(mall pack)-2.75 E(et drop rate, b)-.11 E(ut is allo)-.22 E
(wed to send at the sending rate of a TCP or)-.275 E
(standard TFRC \215o)100.8 111 Q 2.75(wu)-.275 G(sing lar)-2.75 E
(ger pack)-.198 E(ets b)-.11 E(ut recei)-.22 E(ving the same pack)-.275
E(et drop rate.)-.11 E(Simulations sho)100.8 127.6 Q(wing TCP)-.275 E
2.75(,s)-1.221 G
(tandard TFRC, and TFRC-SP sending rates in response to a)-2.75 E
(con\214gured byte drop rate are gi)100.8 140.6 Q -.165(ve)-.275 G 2.75
(ni).165 G 2.75(nA)-2.75 G(ppendix B.2 .)-2.75 E/F1 11/Times-Bold@0 SF
(4.2.)72 166.6 Q/F2 13/Times-Bold@0 SF(Accounting f)5.5 E
(or Header Size)-.325 E F0([RFC3714] mak)100.8 183.2 Q
(es the optimistic assumption that the limitation of the netw)-.11 E
(ork is in)-.11 E
(bandwidth in bytes per second \(Bps\), and not in CPU c)100.8 196.2 Q
(ycles or in pack)-.165 E(ets per second \(pps\).)-.11 E(Ho)100.8 209.2
Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, s).165 H(ome attention must b\
e paid to the load in pps as well as to the load in Bps.).44 E(Ev)100.8
222.2 Q(en aside from the Min Interv)-.165 E(al, TFRC-SP gi)-.275 E
-.165(ve)-.275 G 2.75(st).165 G(he application some incenti)-2.75 E .33
-.165(ve t)-.275 H 2.75(ou).165 G(se)-2.75 E(fe)100.8 235.2 Q(wer b)
-.275 E(ut lar)-.22 E(ger pack)-.198 E(ets, when lar)-.11 E(ger pack)
-.198 E(ets w)-.11 E(ould suf)-.11 E(\214ce, by including the bandwidth)
-.275 E(used by the pack)100.8 248.2 Q(et header in the allo)-.11 E
(wed sending rate.)-.275 E(As an e)100.8 264.8 Q
(xample, a sender using 120-byte pack)-.165 E
(ets needs a TCP-friendly rate of 128 Kbps to)-.11 E
(send 96 Kbps of application data.)100.8 277.8 Q
(This is because the TCP-friendly rate is reduced by a)5.5 E -.11(fa)
100.8 290.8 S
(ctor of s_true/\(s_true + H\) = 120/160, to account for the ef).11 E
(fect of pack)-.275 E(et headers.)-.11 E(If the)5.5 E
(sender suddenly switched to 40-byte data se)100.8 303.8 Q
(gments, the allo)-.165 E(wed sending rate w)-.275 E(ould reduce)-.11 E
(to 64 Kbps of application data; and the use of one-byte data se)100.8
316.8 Q(gments w)-.165 E(ould reduce the)-.11 E(allo)100.8 329.8 Q
(wed sending rate to 3.12 Kbps of application data.)-.275 E(\(In f)5.5 E
(act, the Min Interv)-.11 E(al w)-.275 E(ould)-.11 E(pre)100.8 342.8 Q
-.165(ve)-.275 G(nt senders from achie).165 E
(ving these rates, since applications using TFRC-SP cannot send)-.275 E
(more than 100 pack)100.8 355.8 Q(ets per second.\))-.11 E(Unless it ha\
s a more precise estimate of the header size, TFRC-SP assumes 40 bytes \
for the)100.8 372.4 Q(header size, although the header could be lar)
100.8 385.4 Q(ger \(due to IP options, IPv6, IP tunnels, and)-.198 E
(the lik)100.8 398.4 Q
(e\) or smaller \(due to header compression\) on the wire.)-.11 E
(Requiring the use of the e)5.5 E(xact)-.165 E(header size in bytes w)
100.8 411.4 Q(ould require signi\214cant additional comple)-.11 E(xity)
-.165 E 2.75(,a)-.715 G(nd w)-2.75 E(ould ha)-.11 E .33 -.165(ve l)-.22
H(ittle).165 E(additional bene\214t.)100.8 424.4 Q(TFRC-SP')5.5 E 2.75
(sd)-.605 G(ef)-2.75 E(ault assumption of a 40-byte header is suf)-.11 E
(\214cient to get a)-.275 E(rough estimate of the throughput, and to gi)
100.8 437.4 Q .33 -.165(ve t)-.275 H(he application some incenti).165 E
.33 -.165(ve n)-.275 H(ot to use).165 E(unnecessarily-man)100.8 450.4 Q
2.75(ys)-.165 G(mall pack)-2.75 E 2.75(ets. Because)-.11 F
(we are only aiming at rough f)2.75 E(airness, and at a)-.11 E
(rough incenti)100.8 463.4 Q .33 -.165(ve f)-.275 H
(or applications, the def).165 E
(ault use of a 40-byte header in the calculations of)-.11 E
(the header bandwidth is suf)100.8 476.4 Q
(\214cient for both IPv4 and IPv6.)-.275 E F1(4.3.)72 502.4 Q F2
(The TFRC-SP Min Inter)5.5 E -.13(va)-.13 G(l).13 E F0
(The header size calculation pro)100.8 519 Q(vides an incenti)-.165 E
.33 -.165(ve f)-.275 H(or applications to use fe).165 E(wer)-.275 E 2.75
(,b)-.44 G(ut lar)-2.97 E(ger)-.198 E(,)-.44 E(pack)100.8 532 Q 2.75
(ets. Another)-.11 F(incenti)2.75 E .33 -.165(ve i)-.275 H 2.75(st).165
G(hat when the path limitation is in pps, the application using)-2.75 E
(more small pack)100.8 545 Q(ets is lik)-.11 E(ely to cause higher pack)
-.11 E(et drop rates, and to ha)-.11 E .33 -.165(ve t)-.22 H 2.75(or)
.165 G(educe its)-2.75 E(sending rate accordingly)100.8 558 Q 5.5(.T)
-.715 G(hat is, if the congestion is in terms of pps, then the \215o)
-5.5 E 2.75(ws)-.275 G(ending)-2.75 E(more pps will increase the pack)
100.8 571 Q(et drop rate, and ha)-.11 E .33 -.165(ve t)-.22 H 2.75(oa)
.165 G(djust its sending rate accordingly)-2.75 E(.)-.715 E(Ho)100.8 584
Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, t).165 H
(he increased congestion caused by the use of small pack).44 E
(ets in an en)-.11 E(vironment)-.44 E(limited by pps is e)100.8 597 Q
(xperienced not only by the \215o)-.165 E 2.75(wu)-.275 G
(sing the small pack)-2.75 E(ets, b)-.11 E(ut by all of the)-.22 E
(competing traf)100.8 610 Q(\214c on that congested link.)-.275 E
(These incenti)5.5 E -.165(ve)-.275 G 2.75(sa).165 G(re therefore insuf)
-2.75 E(\214cient to)-.275 E(pro)100.8 623 Q(vide suf)-.165 E
(\214cient protection for pps netw)-.275 E(ork limitations.)-.11 E
(TFRC-SP)100.8 639.6 Q 2.75(,t)-1.221 G(hen, includes a Min Interv)-2.75
E(al of 10 ms.)-.275 E(This pro)5.5 E(vides additional restrictions on)
-.165 E(the use of unnecessarily man)100.8 652.6 Q 2.75(ys)-.165 G
(mall pack)-2.75 E(ets.)-.11 E(One justi\214cation for the Min Interv)
100.8 669.2 Q
(al is the practical one that the applications that currently)-.275 E
-.11(wa)100.8 682.2 S(nt to send small pack).11 E(ets are the V)-.11 E
(oIP applications that send at most one pack)-1.419 E(et e)-.11 E -.165
(ve)-.275 G(ry 10).165 E(ms, so this restriction does not af)100.8 695.2
Q(fect current traf)-.275 E 2.75(\214c. A)-.275 F
(second justi\214cation is that there is)2.75 E(Flo)72 733 Q(yd/K)-.11 E
270.421(ohler Section)-.385 F 2.75(4.3. [P)2.75 F(age 12])-.165 E 0 Cg
EP
%%Page: 13 13
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(no pressing need for best-ef)
100.8 85 Q(fort traf)-.275 E
(\214c in the current Internet to send small pack)-.275 E(ets more)-.11
E(frequently than once e)100.8 98 Q -.165(ve)-.275 G
(ry 10 ms \(rather than taking the 10 ms delay at the sender).165 E 2.75
(,a)-.44 G(nd)-2.75 E(mer)100.8 111 Q(ging the tw)-.198 E 2.75(os)-.11 G
(mall pack)-2.75 E(ets into one lar)-.11 E(ger one\).)-.198 E
(This 10 ms delay for mer)5.5 E(ging small)-.198 E(pack)100.8 124 Q
(ets is lik)-.11 E(ely to be dominated by the netw)-.11 E(ork propag)
-.11 E(ation, transmission, and queueing)-.055 E(delays of best-ef)100.8
137 Q(fort traf)-.275 E(\214c in the current Internet.)-.275 E
(As a result, our judgment w)5.5 E(ould be that)-.11 E
(the bene\214t to the user of ha)100.8 150 Q
(ving less than 10 ms between pack)-.22 E(ets is outweighed by the)-.11
E(bene\214t to the netw)100.8 163 Q(ork of a)-.11 E -.22(vo)-.22 G
(iding unnecessarily man).22 E 2.75(ys)-.165 G(mall pack)-2.75 E(ets.)
-.11 E(The Min Interv)100.8 179.6 Q
(al causes TFRC-SP not to support applications sending small pack)-.275
E(ets v)-.11 E(ery)-.165 E(frequently)100.8 192.6 Q 5.5(.C)-.715 G
(onsider a TFRC \215o)-5.5 E 2.75(ww)-.275 G(ith a \214x)-2.75 E
(ed pack)-.165 E(et size of 100 bytes, b)-.11 E(ut with a v)-.22 E
(ariable)-.275 E(sending rate and a f)100.8 205.6 Q
(airly uncongested path.)-.11 E(When this \215o)5.5 E 2.75(ww)-.275 G
(as sending at most 100 pps, it)-2.86 E -.11(wo)100.8 218.6 S
(uld be able to use TFRC-SP).11 E 5.5(.I)-1.221 G 2.75(ft)-5.5 G
(he \215o)-2.75 E 2.75(ww)-.275 G
(ished to increase its sending rate to more than)-2.75 E(100 pps, b)
100.8 231.6 Q(ut to k)-.22 E(eep the same pack)-.11 E(et size, it w)-.11
E(ould no longer be able to achie)-.11 E .33 -.165(ve t)-.275 H
(his with).165 E(TFRC-SP)100.8 244.6 Q 2.75(,a)-1.221 G(nd w)-2.75 E
(ould ha)-.11 E .33 -.165(ve t)-.22 H 2.75(os).165 G(witch to the def)
-2.75 E(ault TFRC, recei)-.11 E(ving a dramatic,)-.275 E
(discontinuous decrease in its allo)100.8 257.6 Q(wed sending rate.)
-.275 E(This seems not only acceptable, b)5.5 E(ut)-.22 E
(desirable for the global Internet.)100.8 270.6 Q(What is to pre)100.8
287.2 Q -.165(ve)-.275 G(nt \215o).165 E
(ws from opening multiple connections, each with a 10 ms Min)-.275 E
(Interv)100.8 300.2 Q
(al, and thereby getting around the limitation of the Min Interv)-.275 E
2.75(al? Ob)-.275 F(viously)-.165 E 2.75(,t)-.715 G(here is)-2.75 E
(nothing to pre)100.8 313.2 Q -.165(ve)-.275 G(nt \215o).165 E
(ws from doing this, just as there is currently nothing to pre)-.275 E
-.165(ve)-.275 G(nt \215o).165 E(ws)-.275 E(from using UDP)100.8 326.2 Q
2.75(,o)-1.221 G 2.75(rf)-2.75 G
(rom opening multiple parallel TCP connections, or from using their)
-2.75 E -.275(ow)100.8 339.2 S 2.75(nc).275 G
(ongestion control mechanism.)-2.75 E
(Of course, implementations or middlebox)5.5 E(es are also)-.165 E(free\
 to limit the number of parallel TFRC connections opened to the same de\
stination in)100.8 352.2 Q
(times of congestion, if that seems desirable.)100.8 365.2 Q(And \215o)
5.5 E(ws that open multiple parallel)-.275 E
(connections are subject to the incon)100.8 378.2 Q -.165(ve)-.44 G
(niences of reordering and the lik).165 E(e.)-.11 E/F1 11/Times-Bold@0
SF(4.4.)72 404.2 Q/F2 13/Times-Bold@0 SF(Counting P)5.5 E(ack)-.13 E
(et Losses)-.13 E F0(It is not possible for a TCP connection to persist\
ently send multiple pack)100.8 420.8 Q(ets per round-trip)-.11 E
(time in the f)100.8 433.8 Q
(ace of high congestion, with a steady-state with multiple pack)-.11 E
(ets dropped per)-.11 E(round-trip time.)100.8 446.8 Q -.165(Fo)5.5 G
2.75(rT).165 G(CP)-2.75 E 2.75(,w)-1.221 G(hen one or more pack)-2.75 E
(ets are dropped each round-trip time, the)-.11 E
(sending rate is quickly dropped to less than one pack)100.8 459.8 Q
(et per round-trip time.)-.11 E(In addition, for)5.5 E(TCP with T)100.8
472.8 Q(ahoe, Ne)-.88 E(wReno, or SA)-.275 E
(CK congestion control mechanisms, the response to)-.44 E
(congestion is lar)100.8 485.8 Q(gely independent of the number of pack)
-.198 E(ets dropped per round-trip time.)-.11 E
(As a result, standard TFRC can best achie)100.8 502.4 Q .33 -.165(ve f)
-.275 H(airness with TCP).055 E 2.75(,e)-1.221 G -.165(ve)-3.025 G 2.75
(ni).165 G 2.75(nh)-2.75 G(ighly congested)-2.75 E(en)100.8 515.4 Q
(vironments, by calculating the loss e)-.44 E -.165(ve)-.275 G
(nt rate rather than the pack).165 E(et drop rate, where a)-.11 E
(loss e)100.8 528.4 Q -.165(ve)-.275 G(nt is one or more pack).165 E
(ets dropped or mark)-.11 E(ed from a windo)-.11 E 2.75(wo)-.275 G 2.75
(fd)-2.75 G(ata.)-2.75 E(Ho)100.8 545 Q(we)-.275 E -.165(ve)-.275 G .88
-.44(r, w).165 H(ith TFRC-SP).44 E 2.75(,i)-1.221 G 2.75(ti)-2.75 G 2.75
(sn)-2.75 G 2.75(ol)-2.75 G(onger possible to achie)-2.75 E .33 -.165
(ve f)-.275 H(airness with TCP or with).055 E
(standard TFRC by counting only the loss e)100.8 558 Q -.165(ve)-.275 G
(nt rate [WBL04].).165 E(Instead of sending one)5.5 E(lar)100.8 571 Q
(ge pack)-.198 E
(et per round-trip time, TFRC-SP could be sending N small pack)-.11 E
(ets \(where N)-.11 E(small pack)100.8 584 Q(ets equal one lar)-.11 E
(ge 1500-byte pack)-.198 E 2.75(et\). The)-.11 F
(loss measurement used with TFRC-)2.75 E
(SP has to be able to detect a connection that is consistently recei)
100.8 597 Q(ving multiple pack)-.275 E(et losses)-.11 E
(or marks per round-trip time, to allo)100.8 610 Q 2.75(wT)-.275 G
(FRC-SP to respond appropriately)-2.75 E(.)-.715 E(In TFRC-SP)100.8
626.6 Q 2.75(,t)-1.221 G(he loss e)-2.75 E -.165(ve)-.275 G
(nt rate is calculated by counting at most one loss e).165 E -.165(ve)
-.275 G(nt in loss).165 E(interv)100.8 639.6 Q(als longer than tw)-.275
E 2.75(or)-.11 G(ound-trip times, and by counting each pack)-2.75 E
(et lost or mark)-.11 E(ed in)-.11 E(shorter loss interv)100.8 652.6 Q
2.75(als. In)-.275 F(particular)2.75 E 2.75(,f)-.44 G
(or a short loss interv)-2.75 E(al with N pack)-.275 E(ets, including K)
-.11 E(lost or mark)100.8 665.6 Q(ed pack)-.11 E(ets, the loss interv)
-.11 E(al length is calculated as N/K, instead as N.)-.275 E(The)5.5 E
-2.475 -.22(av e)100.8 678.6 T(rage loss interv).22 E
(al I_mean is still a)-.275 E -.165(ve)-.22 G(raged o).165 E -.165(ve)
-.165 G 2.75(rt).165 G(he most recent eight loss interv)-2.75 E(als, as)
-.275 E(speci\214ed in Section 5.4 of RFC 3448.)100.8 691.6 Q
(Thus, if eight successi)5.5 E .33 -.165(ve l)-.275 H(oss interv).165 E
(als are short loss)-.275 E(Flo)72 733 Q(yd/K)-.11 E 270.421
(ohler Section)-.385 F 2.75(4.4. [P)2.75 F(age 13])-.165 E 0 Cg EP
%%Page: 14 14
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(interv)100.8 85 Q
(als with N pack)-.275 E(ets and K losses, the loss e)-.11 E -.165(ve)
-.275 G(nt rate is calculated as K/N, rather than).165 E(as 1/N.)100.8
98 Q(TFRC-SP follo)100.8 114.6 Q
(ws the procedures from [RFC3448] in terms of)-.275 E/F1 11/Times-Bold@0
SF(4.5.)72 140.6 Q/F2 13/Times-Bold@0 SF(The Nominal P)5.5 E(ack)-.13 E
(et Size)-.13 E F0(The guidelines in Section 3 abo)100.8 157.2 Q .33
-.165(ve s)-.165 H(ay that the nominal se).165 E
(gment size s is set to 1460 bytes,)-.165 E(pro)100.8 170.2 Q
(viding a goal of f)-.165 E
(airness, in terms of the sending rate in bytes per second, with a TCP)
-.11 E<8d6f>100.8 183.2 Q 2.75(ww)-.275 G
(ith 1460 bytes of application data per pack)-2.75 E(et b)-.11 E
(ut with the same pack)-.22 E(et drop rate.)-.11 E(This follo)100.8
196.2 Q(ws the assumption that a TCP \215o)-.275 E 2.75(ww)-.275 G
(ith 1460-byte se)-2.75 E(gments will ha)-.165 E .33 -.165(ve a h)-.22 H
(igher).165 E(sending rate than a TCP \215o)100.8 209.2 Q 2.75(ww)-.275
G(ith smaller se)-2.75 E 2.75(gments. While)-.165 F
(this assumption holds in an)2.75 E(en)100.8 222.2 Q
(vironment where the pack)-.44 E(et drop rate is independent of pack)
-.11 E(et size, this assumption does)-.11 E
(not necessarily hold in an en)100.8 235.2 Q(vironment where lar)-.44 E
(ger pack)-.198 E(ets are more lik)-.11 E(ely to be dropped)-.11 E
(than are small pack)100.8 248.2 Q(ets.)-.11 E(The table belo)100.8
264.8 Q 2.75(ws)-.275 G(ho)-2.75 E
(ws the results of simulations with standard \(SA)-.275 E
(CK\) TCP \215o)-.44 E(ws, where,)-.275 E(for each *byte*, the pack)
100.8 277.8 Q(et containing that byte is dropped with probability p.)
-.11 E(Consider)5.5 E
(the approximation for the TCP response function for pack)100.8 290.8 Q
(et drop rates up to 10% or so; for)-.11 E(this en)100.8 303.8 Q
(vironments, the sending rate in bytes per R)-.44 E
(TT is roughly 1.2 s/sqrt\(p\), for a pack)-.66 E(et)-.11 E
(size of s bytes and pack)100.8 316.8 Q(et drop rate p.)-.11 E(Gi)5.5 E
-.165(ve)-.275 G 2.75(na\214).165 G -.165(xe)-2.75 G 2.75(d*).165 G
(byte* drop rate p1, and a TCP pack)-2.75 E(et)-.11 E
(size of s bytes, the pack)100.8 329.8 Q
(et drop rate is roughly s*p1, producing a sending rate in bytes per)
-.11 E -.66(RT)100.8 342.8 S 2.75(To).66 G 2.75(fr)-2.75 G
(oughly 1.2 sqrt\(s\)/sqrt\(p1\).)-2.75 E(Thus, for TCP in an en)5.5 E
(vironment with a \214x)-.44 E(ed byte)-.165 E
(drop rate, the sending rate should gro)100.8 355.8 Q 2.75(wr)-.275 G
(oughly as sqrt\(s\), for pack)-2.75 E(et drop rates up to 10%)-.11 E
(or so.)100.8 368.8 Q(Each ro)100.8 385.4 Q 2.75(wo)-.275 G 2.75(fT)
-2.75 G(able 5 belo)-3.63 E 2.75(ws)-.275 G(ho)-2.75 E
(ws a separate simulation with ten TCP connection and a)-.275 E<8c78>
100.8 398.4 Q
(ed byte drop rate of 0.0001, with each simulation using a dif)-.165 E
(ferent se)-.275 E(gment size.)-.165 E -.165(Fo)5.5 G(r).165 E
(each simulation, the TCP sending rate and goodput are a)100.8 411.4 Q
-.165(ve)-.22 G(raged o).165 E -.165(ve)-.165 G 2.75(rt).165 G
(he ten \215o)-2.75 E 2.75(ws. As)-.275 F(one)2.75 E -.11(wo)100.8 424.4
S(uld e).11 E(xpect from the paragraph abo)-.165 E -.165(ve)-.165 G 2.75
(,t).165 G(he TCP sending rate gro)-2.75 E(ws some)-.275 E
(what less than)-.275 E(linearly with an increase in pack)100.8 437.4 Q
(et size, up to a pack)-.11 E(et size of 1460 bytes, corresponding to)
-.11 E 2.75(ap)100.8 450.4 S(ack)-2.75 E(et drop rate of 13%.)-.11 E
(After that, further increases in the pack)5.5 E(et size result in a)
-.11 E(*decrease* in the TCP sending rate, as the TCP connection enters\
 the re)100.8 463.4 Q(gime of)-.165 E -.165(ex)100.8 476.4 S
(ponential back).165 E(of)-.11 E 2.75(fo)-.275 G 2.75(ft)-2.75 G
(he retransmit timer)-2.75 E(.)-.605 E/F3 11/Courier@0 SF 13.2
(Segment Packet)193.2 495.4 R(TCP Rates \(Kbps\))39.6 E(Size \(B\))193.2
508.4 Q 13.2(DropRate SendRate)13.2 F(Goodput)26.4 E 6.6
(-------- --------)193.2 521.4 R 19.8(-------- -------)19.8 F 33
(14 0.005)219.6 534.4 R 39.6(6.37 6.34)46.2 F 33(128 0.016 30.78 30.30)
213 547.4 R 33(256 0.028 46.54 44.96)213 560.4 R 33
(512 0.053 62.43 58.37)213 573.4 R 33(1460 0.134 94.15 80.02)206.4 586.4
R 33(4000 0.324 35.20 21.44)206.4 599.4 R 33(8000 0.531 15.36)206.4
612.4 R(5.76)46.2 E(Table 5: TCP Median Send Rate vs. Packet Size I:)
173.4 638.4 Q(Byte Drop Rate 0.0001)259.2 651.4 Q F0 -.88(Ta)100.8 670.4
S(ble 6 belo).88 E 2.75(ws)-.275 G(ho)-2.75 E
(ws similar results for a byte drop rate of 0.001.)-.275 E
(In this case, the TCP)5.5 E(sending rate gro)100.8 683.4 Q
(ws with increasing pack)-.275 E(et size up to a pack)-.11 E
(et size of 128 bytes,)-.11 E(corresponding to a pack)100.8 696.4 Q
(et drop rate of 16%.)-.11 E(After than, the TCP sending rate decreases)
5.5 E(Flo)72 733 Q(yd/K)-.11 E 270.421(ohler Section)-.385 F 2.75
(4.5. [P)2.75 F(age 14])-.165 E 0 Cg EP
%%Page: 15 15
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(and then increases ag)100.8 85
Q(ain, as the TCP connection enters the re)-.055 E(gime of e)-.165 E
(xponential back)-.165 E(of)-.11 E(f)-.275 E(of the retransmit timer)
100.8 98 Q 5.5(.N)-.605 G(ote that with this byte drop rate, with pack)
-5.5 E(et sizes of 4000 and)-.11 E
(8000 bytes, the TCP sending rate increases b)100.8 111 Q
(ut the TCP goodput rate remains essentially)-.22 E 2.75(zero. This)
100.8 124 R(mak)2.75 E(es sense, as almost all pack)-.11 E
(ets that are sent are dropped.)-.11 E/F1 11/Courier@0 SF 13.2
(Segment Packet)193.2 143 R(TCP Rates \(Kbps\))39.6 E(Size \(B\))193.2
156 Q 13.2(DropRate SendRate)13.2 F(Goodput)26.4 E 6.6
(-------- --------)193.2 169 R 19.8(-------- -------)19.8 F 33(14 0.053)
219.6 182 R 39.6(1.68 1.56)46.2 F 33(128 0.159)213 195 R 39.6(7.66 6.13)
46.2 F 33(256 0.248)213 208 R 39.6(6.21 4.32)46.2 F 33(512 0.402)213 221
R 39.6(1.84 1.11)46.2 F 33(1460 0.712)206.4 234 R 39.6(1.87 0.47)46.2 F
33(4000 0.870)206.4 247 R 39.6(3.20 0.00)46.2 F 33(8000 0.890)206.4 260
R 39.6(5.76 0.00)46.2 F
(Table 6: TCP Median Send Rate vs. Packet Size II:)173.4 286 Q
(Byte Drop Rate 0.001)259.2 299 Q F0(The TCP beha)100.8 318 Q
(vior in the presence of a \214x)-.22 E
(ed byte drop rate suggests that instead of the goal)-.165 E
(of a TFRC-SP \215o)100.8 331 Q 2.75(wa)-.275 G(chie)-2.75 E
(ving the same sending rate in bytes per second as a TCP \215o)-.275 E
(w)-.275 E(using 1500-byte pack)100.8 344 Q(ets, it mak)-.11 E
(es more sense to consider an ideal goal of a TFRC-SP \215o)-.11 E(w)
-.275 E(achie)100.8 357 Q(ving the same sending rate as a TCP \215o)
-.275 E 2.75(ww)-.275 G(ith the optimal pack)-2.75 E(et size, gi)-.11 E
-.165(ve)-.275 G 2.75(nt).165 G(hat the)-2.75 E(pack)100.8 370 Q
(et size is at most 1500 bytes.)-.11 E
(This does not mean that we need to change the TFRC-)5.5 E
(SP mechanisms for computing the allo)100.8 383 Q(wed transmit rate;)
-.275 E(this means simply that in)5.5 E -.275(eva)100.8 396 S
(luating the f).275 E(airness of TFRC-SP)-.11 E 2.75(,w)-1.221 G 2.75
(es)-2.75 G(hould consider f)-2.75 E(airness relati)-.11 E .33 -.165
(ve t)-.275 H 2.75(ot).165 G(he TCP \215o)-2.75 E(w)-.275 E
(using the optimal pack)100.8 409 Q
(et size \(though still at most 1500 bytes\) for that en)-.11 E
(vironment.)-.44 E/F2 11/Times-Bold@0 SF(4.6.)72 435 Q/F3 13
/Times-Bold@0 SF(The Calculated Loss Inter)5.5 E -.13(va)-.13 G 3.25(lL)
.13 G(ength f)-3.25 E(or Short Loss Inter)-.325 E -.13(va)-.13 G(ls).13
E F0 -.165(Fo)100.8 451.6 S 2.75(raT).165 G(FRC-SP recei)-2.75 E -.165
(ve)-.275 G .88 -.44(r, t).165 H
(he guidelines from Section 6 of RFC 3448 go).44 E -.165(ve)-.165 G
(rn when the).165 E(recei)100.8 464.6 Q -.165(ve)-.275 G 2.75(rs).165 G
(hould send feedback messages.)-2.75 E(In particular)5.5 E 2.75(,f)-.44
G(rom [RFC3448], "a feedback)-2.75 E(pack)100.8 477.6 Q
(et should ... be sent whene)-.11 E -.165(ve)-.275 G 2.75(ran).165 G .55
-.275(ew l)-2.75 H(oss e).275 E -.165(ve)-.275 G
(nt is detected without w).165 E(aiting for the end)-.11 E(of an R)100.8
490.6 Q 2.75(TT". In)-.66 F(addition, feedback pack)2.75 E
(ets are sent at least once per R)-.11 E(TT)-.66 E(.)-.814 E 1.76 -.88
(We n)100.8 507.2 T(ote that in a connection with a short loss interv)
.88 E(al \(less that tw)-.275 E 2.75(or)-.11 G(ound-trip times\) and)
-2.75 E(multiple pack)100.8 520.2 Q(et losses per loss interv)-.11 E
(al, the loss interv)-.275 E(al length calculated for the most)-.275 E
(recent loss interv)100.8 533.2 Q
(al can decrease from one round-trip time to the ne)-.275 E
(xt, as multiple pack)-.165 E(et)-.11 E
(losses are detected for that loss interv)100.8 546.2 Q 2.75(al. As)
-.275 F(an e)2.75 E(xample, when the TFRC-SP recei)-.165 E -.165(ve)
-.275 G 2.75(rs).165 G(ends)-2.75 E 2.75(af)100.8 559.2 S(eedback pack)
-2.75 E(et the current loss interv)-.11 E(al might contain N pack)-.275
E(ets, with only one loss,)-.11 E(gi)100.8 572.2 Q
(ving a calculated loss interv)-.275 E(al length for that interv)-.275 E
(al of N pack)-.275 E 2.75(ets. When)-.11 F(the TFRC-SP)2.75 E(recei)
100.8 585.2 Q -.165(ve)-.275 G 2.75(rs).165 G(ends a feedback pack)-2.75
E(et one round-trip time later)-.11 E 2.75(,Ka)-.44 G
(dditional lost or mark)-2.75 E(ed)-.11 E(pack)100.8 598.2 Q
(ets might ha)-.11 E .33 -.165(ve b)-.22 H(een detected, gi).165 E
(ving a calculated loss interv)-.275 E(al length for that interv)-.275 E
(al)-.275 E(of only \(N+K\)/\(K+1\) pack)100.8 611.2 Q 2.75(ets. F)-.11
F(or K=N/2, this could lead to a change in the calculated)-.165 E
(loss interv)100.8 624.2 Q(al length from N to close to 2 pack)-.275 E
2.75(ets. T)-.11 F 2.75(op)-.88 G(re)-2.75 E -.165(ve)-.275 G
(nt unnecessary oscillations in the).165 E -2.475 -.22(av e)100.8 637.2
T(rage loss interv).22 E
(al, Section 3 speci\214es that the current loss interv)-.275 E
(al can be included in the)-.275 E(calculation of the a)100.8 650.2 Q
-.165(ve)-.22 G(rage loss interv).165 E
(al only if the current loss interv)-.275 E(al is longer than tw)-.275 E
(o)-.11 E(round-trip times.)100.8 663.2 Q -.165(Fo)5.5 G 2.75(ral).165 G
(oss interv)-2.75 E(al longer than tw)-.275 E 2.75(or)-.11 G
(ound-trip times, the detection of ne)-2.75 E(w)-.275 E
(losses for the loss interv)100.8 676.2 Q
(al will not *decrease* the calculated loss interv)-.275 E
(al length for that)-.275 E(loss interv)100.8 689.2 Q(al.)-.275 E(Flo)72
733 Q(yd/K)-.11 E 270.421(ohler Section)-.385 F 2.75(4.6. [P)2.75 F
(age 15])-.165 E 0 Cg EP
%%Page: 16 16
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Times-Bold@0 SF(5.)72 85
Q/F2 14/Times-Bold@0 SF 3.5(AC)5.5 G(omparison with RFC 3714)-3.5 E F0
(RFC 3714 considers the problems of f)100.8 101.6 Q
(airness, potential congestion collapse, and poor user)-.11 E
(quality that could occur with the deplo)100.8 114.6 Q
(yment of non-congestion-controlled IP telephon)-.11 E(y)-.165 E -.165
(ove)100.8 127.6 S 2.75(rc).165 G(ongested best-ef)-2.75 E(fort netw)
-.275 E 2.75(orks. The)-.11 F(March 2004 document cites ongoing ef)2.75
E(forts in)-.275 E(the IETF)100.8 140.6 Q 2.75(,i)-.88 G(ncluding w)
-2.75 E(ork on TFRC and DCCP)-.11 E 5.5(.R)-1.221 G
(FC 3714 recommends that for best-ef)-5.5 E(fort)-.275 E(traf)100.8
153.6 Q(\214c with applications that ha)-.275 E .33 -.165(ve a \214)-.22
H -.165(xe).165 G 2.75(do).165 G 2.75(rm)-2.75 G
(inimum sending rate, the application or)-2.75 E
(transport protocol should monitor the pack)100.8 166.6 Q
(et drop rate, and discontinue sending for a)-.11 E
(period if the steady-state pack)100.8 179.6 Q
(et drop rate signi\214cantly e)-.11 E(xceeds the allo)-.165 E
(wed threshold for)-.275 E(that minimum sending rate.)100.8 192.6 Q
(In determining the allo)100.8 209.2 Q(wed pack)-.275 E
(et drop rate for a \214x)-.11 E(ed sending rate, RFC 3714 assumes)-.165
E(that an IP telephon)100.8 222.2 Q 2.75<798d>-.165 G .55 -.275(ow s)
-2.75 H(hould be allo).275 E
(wed to use the same sending rate in bytes per)-.275 E
(second as a 1500-byte-pack)100.8 235.2 Q(et TCP \215o)-.11 E 2.75(we)
-.275 G(xperiencing the same pack)-2.915 E(et drop rate as that of)-.11
E(the IP telephon)100.8 248.2 Q 2.75<798d>-.165 G -.275(ow)-2.75 G 5.5
(.A)-.44 G 2.75(sa)-5.5 G 2.75(ne)-2.75 G(xample, follo)-2.915 E
(wing this guideline, a V)-.275 E(oIP connection with a)-1.419 E
(round-trip time of 0.1 sec and a minimum sending rate of 64 Kbps w)
100.8 261.2 Q(ould be required to)-.11 E
(terminate or suspend when the persistent pack)100.8 274.2 Q
(et drop rate signi\214cantly e)-.11 E(xceeded 25%.)-.165 E(One limitat\
ion of the lack of \214ne-grained control in the minimal mechanism desc\
ribed in)100.8 290.8 Q(RFC 3714 is that an IP telephon)100.8 303.8 Q
2.75<798d>-.165 G .55 -.275(ow w)-2.75 H
(ould not adapt its sending rate in response to).165 E 2.75
(congestion. In)100.8 316.8 R(contrast, with TFRC-SP)2.75 E 2.75(,as)
-1.221 G(mall-pack)-2.75 E(et \215o)-.11 E 2.75(ww)-.275 G
(ould reduce its sending rate)-2.86 E(some)100.8 329.8 Q
(what in response to moderate pack)-.275 E(et drop rates, possibly a)
-.11 E -.22(vo)-.22 G(iding a period with).22 E(unnecessarily-hea)100.8
342.8 Q(vy pack)-.22 E(et drop rates in the netw)-.11 E(ork.)-.11 E
(Because RFC 3714 assumes that the allo)100.8 359.4 Q(wed pack)-.275 E
(et drop rate for an IP telephon)-.11 E 2.75<798d>-.165 G .55 -.275
(ow i)-2.75 H(s).275 E(determined by the sending rate that a TCP \215o)
100.8 372.4 Q 2.75(ww)-.275 G(ould use *with the same pack)-2.86 E
(et drop)-.11 E(rate*, the minimal mechanism in RFC 3714 w)100.8 385.4 Q
(ould not pro)-.11 E(vide f)-.165 E(airness between TCP and)-.11 E
(IP telephon)100.8 398.4 Q 2.75(yt)-.165 G(raf)-2.75 E(\214c in an en)
-.275 E(vironment where small pack)-.44 E(ets are less lik)-.11 E
(ely to be dropped)-.11 E(than lar)100.8 411.4 Q(ge pack)-.198 E 2.75
(ets. In)-.11 F(such an en)2.75 E(vironment, the small-pack)-.44 E
(et IP telephon)-.11 E 2.75<798d>-.165 G .55 -.275(ow w)-2.75 H(ould)
.165 E(mak)100.8 424.4 Q 2.75(et)-.11 G
(he optimistic assumption that a lar)-2.75 E(ge-pack)-.198 E
(et TCP \215o)-.11 E 2.75(ww)-.275 G(ould recei)-2.86 E .33 -.165(ve t)
-.275 H(he same).165 E(pack)100.8 437.4 Q
(et drop rate as the IP telephon)-.11 E 2.75<798d>-.165 G -.275(ow)-2.75
G 2.75(,a)-.44 G(nd as a result the small-pack)-2.75 E(et IP telephon)
-.11 E 2.75<798d>-.165 G -.275(ow)-2.75 G -.11(wo)100.8 450.4 S
(uld recei).11 E .33 -.165(ve a l)-.275 H(ar).165 E
(ger fraction of the link bandwidth than a competing lar)-.198 E
(ge-pack)-.198 E(et TCP)-.11 E<8d6f>100.8 463.4 Q -.715(w.)-.275 G F1
(6.)72 489.4 Q F2(TFRC-SP with A)5.5 E(pplications that Modify the P)
-.35 E(ack)-.14 E(et Size)-.14 E F0(One possible use for TFRC-SP w)100.8
506 Q(ould be with applications that maintain a \214x)-.11 E(ed sending)
-.165 E(rate in pack)100.8 519 Q(ets per second, b)-.11 E
(ut modify their pack)-.22 E(et size in response to congestion.)-.11 E
(TFRC-)5.5 E(SP monitors the connection')100.8 532 Q 2.75(sp)-.605 G
(ack)-2.75 E(et drop rate, and determines the allo)-.11 E
(wed sending rate in)-.275 E(bytes per second.)100.8 545 Q(Gi)5.5 E
-.165(ve)-.275 G 2.75(na).165 G 2.75(na)-2.75 G(pplication with a \214x)
-2.75 E(ed sending rate in pack)-.165 E(ets per second, the)-.11 E
(TFRC-SP sender could determine the data pack)100.8 558 Q
(et size that w)-.11 E(ould be needed for the sending)-.11 E
(rate in bytes per second not to e)100.8 571 Q(xceed the allo)-.165 E
(wed sending rate.)-.275 E(In en)5.5 E(vironments where the)-.44 E(pack)
100.8 584 Q(et drop rate is af)-.11 E(fected by the pack)-.275 E
(et size, decreasing the pack)-.11 E(et size could also result)-.11 E
(in a decrease in the pack)100.8 597 Q(et drop rate e)-.11 E
(xperienced by the \215o)-.165 E -.715(w.)-.275 G(There are man)100.8
613.6 Q 2.75(yq)-.165 G(uestions about ho)-2.75 E 2.75(wa)-.275 G 2.75
(na)-2.75 G(dapti)-2.75 E .33 -.165(ve a)-.275 H(pplication w).165 E
(ould use TFRC-SP that are)-.11 E(be)100.8 626.6 Q
(yond the scope of this document.)-.165 E(In particular)5.5 E 2.75(,a)
-.44 G 2.75(na)-2.75 G(pplication might wish to a)-2.75 E -.22(vo)-.22 G
(id).22 E(unnecessary reductions in the pack)100.8 639.6 Q(et size.)-.11
E(In this case, an application might w)8.25 E(ait for some)-.11 E
(period of time before reducing the pack)100.8 652.6 Q(et size, to a)
-.11 E -.22(vo)-.22 G(id an unnecessary reduction in the).22 E(pack)
100.8 665.6 Q(et size.)-.11 E(The details of ho)5.5 E 2.75(wl)-.275 G
(ong an application might w)-2.75 E(ait before reducing the pack)-.11 E
(et)-.11 E(size can be addressed in documents that are more application\
-speci\214c.)100.8 678.6 Q(Flo)72 733 Q(yd/K)-.11 E 278.671
(ohler Section)-.385 F 2.75(6. [P)2.75 F(age 16])-.165 E 0 Cg EP
%%Page: 17 17
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(Similarly)100.8 85 Q 2.75(,a)
-.715 G 2.75(na)-2.75 G(pplication using TFRC-SP might only ha)-2.75 E
.33 -.165(ve a f)-.22 H .55 -.275(ew p).165 H(ack).275 E
(et sizes that it is able)-.11 E(to use.)100.8 98 Q
(In this case, the application might not reduce the pack)5.5 E
(et size until the current)-.11 E(pack)100.8 111 Q
(et drop rate has signi\214cantly e)-.11 E(xceeded the pack)-.165 E
(et drop rate threshold for the current)-.11 E(sending rate, to a)100.8
124 Q -.22(vo)-.22 G(id unnecessary oscillations between tw).22 E 2.75
(op)-.11 G(ack)-2.75 E(et sizes and tw)-.11 E 2.75(os)-.11 G(ending)
-2.75 E 2.75(rates. Ag)100.8 137 R(ain, the details will ha)-.055 E .33
-.165(ve t)-.22 H 2.75(ob).165 G 2.75(ea)-2.75 G
(ddressed in documents that are more application-)-2.75 E(speci\214c.)
100.8 150 Q/F1 11/Times-Bold@0 SF(7.)72 176 Q/F2 14/Times-Bold@0 SF
(Simulations)5.5 E F0(This section describes the performance of TFRC-SP\
 in simulation scenarios with)100.8 192.6 Q(con\214gured pack)100.8
205.6 Q(et or byte drop rates, and in scenarios with a range of queue m\
anagement)-.11 E(mechanisms at the congested link.)100.8 218.6 Q
(The simulations, described in detail in Appendix B,)5.5 E -.165(ex)
100.8 231.6 S(plore en).165 E(vironments where standard TFRC signi\214c\
antly limits the throughput of small-)-.44 E(pack)100.8 244.6 Q
(et \215o)-.11 E(ws, and TFRC-SP gi)-.275 E -.165(ve)-.275 G 2.75(st)
.165 G(he desired throughput.)-2.75 E(The simulations also e)5.5 E
(xplore)-.165 E(en)100.8 257.6 Q(vironments where standard TFRC allo)
-.44 E(ws small-pack)-.275 E(et \215o)-.11 E(ws to recei)-.275 E .33
-.165(ve g)-.275 H(ood).165 E(performance, while TFRC-SP is o)100.8
270.6 Q -.165(ve)-.165 G(rly aggressi).165 E -.165(ve)-.275 G(.).165 E
(The general lessons from the simulations are as follo)100.8 287.2 Q
(ws.)-.275 E 12.65<8349>100.8 303.8 S 2.75(ns)-12.65 G
(cenarios where lar)-2.75 E(ge and small pack)-.198 E(ets recei)-.11 E
.33 -.165(ve s)-.275 H(imilar pack).165 E(et drop rates, TFRC-SP)-.11 E
(gi)117.3 316.8 Q -.165(ve)-.275 G 2.75(sr).165 G
(oughly the desired sending rate \(Appendix B.1, B.2\).)-2.75 E 12.65
<8349>100.8 333.4 S 2.75(ns)-12.65 G
(cenarios where each *byte* is equally lik)-2.75 E
(ely to be dropped, standard TFRC gi)-.11 E -.165(ve)-.275 G(s).165 E
(reasonable f)117.3 346.4 Q(airness between small-pack)-.11 E
(et TFRC \215o)-.11 E(ws and lar)-.275 E(ge-pack)-.198 E(et TCP \215o)
-.11 E(ws)-.275 E(\(Appendix B.2\).)117.3 359.4 Q 12.65<8349>100.8 376 S
2.75(ns)-12.65 G(cenarios where small pack)-2.75 E(ets are less lik)-.11
E(ely to be dropped than lar)-.11 E(ge pack)-.198 E(ets,)-.11 E
(TFRC-SP does not gi)117.3 389 Q .33 -.165(ve r)-.275 H(easonable f).165
E(airness between small-pack)-.11 E(et TFRC-SP \215o)-.11 E(ws and)-.275
E(lar)117.3 402 Q(ge-pack)-.198 E(et TCP \215o)-.11 E(ws; small-pack)
-.275 E(et TFRC-SP \215o)-.11 E(ws can recei)-.275 E .33 -.165(ve c)
-.275 H(onsiderably more).165 E(bandwidth than competing lar)117.3 415 Q
(ge-pack)-.198 E(et TCP \215o)-.11 E(ws, and in some cases lar)-.275 E
(ge-pack)-.198 E(et)-.11 E(TCP \215o)117.3 428 Q
(ws can be essentially starv)-.275 E(ed by competing small-pack)-.165 E
(et TFRC-SP \215o)-.11 E(ws.)-.275 E(\(Appendix B.2, B.3, B.4\).)117.3
441 Q 12.65<8353>100.8 457.6 S(cenarios where small pack)-12.65 E
(ets are less lik)-.11 E(ely to be dropped than lar)-.11 E(ge pack)-.198
E(ets include)-.11 E(those with Drop-T)117.3 470.6 Q
(ail queues in bytes, and with A)-.88 E(QM mechanisms in byte mode)-.605
E(\(Appendix B.3, B.4\).)117.3 483.6 Q
(It has also been reported that wireless links are sometimes good)5.5 E
(enough to let small pack)117.3 496.6 Q(ets through, while lar)-.11 E
(ger pack)-.198 E(ets ha)-.11 E .33 -.165(ve a m)-.22 H
(uch higher error rate,).165 E
(and hence a higher drop probability [S05].)117.3 509.6 Q F1(8.)72 535.6
Q F2(General Discussion)5.5 E F0(Dropping rates for small pack)100.8
552.2 Q(ets: The goal of TFRC-SP is for small-pack)-.11 E
(et TFRC-SP \215o)-.11 E(ws)-.275 E(to ha)100.8 565.2 Q .33 -.165(ve r)
-.22 H(ough f).165 E(airness with lar)-.11 E(ge-pack)-.198 E
(et TCP \215o)-.11 E(ws in the sending rate in bps, in a scenario)-.275
E(where each pack)100.8 578.2 Q(et recei)-.11 E -.165(ve)-.275 G 2.75
(sr).165 G(oughly the same probability of being dropped.)-2.75 E
(In a scenario)5.5 E(where lar)100.8 591.2 Q(ge pack)-.198 E
(ets are more lik)-.11 E(ely to be dropped than small pack)-.11 E
(ets, or where \215o)-.11 E(ws with)-.275 E 2.75(ab)100.8 604.2 S
(ursty sending rate are more lik)-2.97 E(ely to ha)-.11 E .33 -.165
(ve p)-.22 H(ack).165 E(ets dropped than are \215o)-.11 E
(ws with a smooth)-.275 E(sending rate, small-pack)100.8 617.2 Q
(et TFRC-SP \215o)-.11 E(ws can recei)-.275 E .33 -.165(ve s)-.275 H
(igni\214cantly more bandwidth than).165 E(competing lar)100.8 630.2 Q
(ge-pack)-.198 E(et TCP \215o)-.11 E(ws.)-.275 E(The accurac)100.8 646.8
Q 2.75(yo)-.165 G 2.75(ft)-2.75 G
(he TCP response function used in TFRC: F)-2.75 E
(or applications with a)-.165 E
(maximum sending rate of 96 Kbps or less \(i.e., pack)100.8 659.8 Q
(ets of at most 120 bytes\) TFRC-SP)-.11 E
(only restricts the sending rate when the pack)100.8 672.8 Q
(et drop rate is f)-.11 E(airly high, e.g., greater than)-.11 E 2.75
(10%. [Deri)100.8 685.8 R -.275(va)-.275 G(tion: A TFRC-SP \215o).275 E
2.75(ww)-.275 G(ith a 200 ms round-trip time and a maximum)-2.75 E(Flo)
72 733 Q(yd/K)-.11 E 278.671(ohler Section)-.385 F 2.75(8. [P)2.75 F
(age 17])-.165 E 0 Cg EP
%%Page: 18 18
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(sending rate with pack)100.8
85 Q(et headers of 128 Kbps w)-.11 E(ould ha)-.11 E .33 -.165(ve a s)
-.22 H(ending rate in bytes per).165 E(second equi)100.8 98 Q -.275(va)
-.275 G(lent to a TCP \215o).275 E 2.75(ww)-.275 G(ith 1460-byte se)
-2.75 E(gments sending 2.2 pack)-.165 E(ets per round-)-.11 E
(trip time.)100.8 111 Q(From T)5.5 E
(able 1 of RFC 3714, this sending rate can be sustained with a pack)-.88
E(et drop)-.11 E(rate slightly greater than 10%.])100.8 124 Q
(In this high-pack)5.5 E(et-drop re)-.11 E(gime, the performance of)
-.165 E(TFRC-SP is determined in part by the accurac)100.8 137 Q 2.75
(yo)-.165 G 2.75(ft)-2.75 G(he TCP response function in)-2.75 E
(representing the actual sending rate of a TCP connection.)100.8 150 Q
(In the re)100.8 166.6 Q(gime of high pack)-.165 E
(et drop rates, TCP performance is also af)-.11 E(fected by the TCP)
-.275 E(algorithm \(e.g., SA)100.8 179.6 Q
(CK or not\), by the minimum R)-.44 E -.198(TO)-.66 G 2.75(,b).198 G
2.75(yt)-2.75 G(he use or not of Limited)-2.75 E -.385(Tr)100.8 192.6 S
(ansmit, by the use of ECN, and the lik).385 E 2.75(e. It)-.11 F
(is good to ensure that simulations or)2.75 E -.165(ex)100.8 205.6 S
(periments e).165 E(xploring f)-.165 E(airness include the e)-.11 E
(xploration of f)-.165 E(airness with the most aggressi)-.11 E -.165(ve)
-.275 G(TCP mechanisms conformance with the current standards.)100.8
218.6 Q(Our simulations use SA)5.5 E(CK)-.44 E(TCP with Limited T)100.8
231.6 Q(ransmit and with a minimum R)-.385 E .396 -.198(TO o)-.66 H 2.75
(f2).198 G(00 ms.)-2.75 E(The simulation results)5.5 E(are lar)100.8
244.6 Q(gely the same with or without timestamps; timestamps were not u\
sed for simulations)-.198 E(reported in this paper)100.8 257.6 Q 5.5(.W)
-.605 G 2.75(ed)-6.38 G(idn')-2.75 E 2.75(tu)-.198 G
(se TCP with ECN in setting the tar)-2.75 E(get sending rate for)-.198 E
(TFRC, because, as e)100.8 270.6 Q(xplained in Appendix B.1, our e)-.165
E(xpectation is that in high pack)-.165 E(et drop)-.11 E(re)100.8 283.6
Q(gimes, routers will drop rather than mark pack)-.165 E
(ets, either from polic)-.11 E 2.75(yo)-.165 G 2.75(rf)-2.75 G(rom b)
-2.75 E(uf)-.22 E(fer)-.275 E -.165(ove)100.8 296.6 S(r\215o).165 E
-.715(w.)-.275 G -.165(Fa)100.8 313.2 S(irness with dif).165 E
(ferent pack)-.275 E(et header sizes: In en)-.11 E
(vironment with IPv6 and/or se)-.44 E -.165(ve)-.275 G(ral layers).165 E
(of netw)100.8 326.2 Q(ork-layer tunnels \(e.g., IPsec, GRE\), the pack)
-.11 E(et header could be 60, 80, or 100)-.11 E
(bytes instead of the def)100.8 339.2 Q
(ault 40 bytes assumed in Section 3.)-.11 E -.165(Fo)5.5 G 2.75(ra).165
G 2.75(na)-2.75 G(pplication with small)-2.75 E(ten-byte data se)100.8
352.2 Q(gments, this means that the actual pack)-.165 E
(et size could be 70, 90, or 110 bytes,)-.11 E
(instead of the 50 bytes assumed by TFRC-SP in calculating the allo)
100.8 365.2 Q(wed sending rate.)-.275 E
(Thus, a TFRC-SP application with lar)100.8 378.2 Q
(ge headers could recei)-.198 E .33 -.165(ve m)-.275 H
(ore than twice the).165 E(bandwidth \(including the bandwidth used by \
headers\) than a TFRC-SP application with)100.8 391.2 Q(small headers.)
100.8 404.2 Q 1.76 -.88(We d)5.5 H 2.75(on).88 G(ot e)-2.75 E
(xpect this to be a problem; in particular)-.165 E 2.75(,T)-.44 G
(FRC-SP applications)-2.75 E(with lar)100.8 417.2 Q
(ge headers will not signi\214cantly de)-.198 E
(grade the performance of competing TCP)-.165 E
(applications, or of competing TFRC-SP applications with small headers.)
100.8 430.2 Q(General issues for TFRC: The congestion control mechanism\
s in TFRC and TFRC-SP)100.8 446.8 Q(limit a \215o)100.8 459.8 Q(w')-.275
E 2.75(ss)-.605 G(ending rate in pack)-2.75 E(ets per second.)-.11 E
(Simulations by T)5.5 E(om Phelan [P04])-.88 E -.165(ex)100.8 472.8 S
(plore ho).165 E 2.75(ws)-.275 G
(uch a limitation in sending rate can lead to unf)-2.75 E
(airness for the TFRC \215o)-.11 E 2.75(wi)-.275 G(n)-2.75 E
(some scenarios, e.g., when competing with b)100.8 485.8 Q
(ursty TCP web traf)-.22 E(\214c, in scenarios with lo)-.275 E(w)-.275 E
(le)100.8 498.8 Q -.165(ve)-.275 G(ls of statistical multiple).165 E
(xing at the congested link.)-.165 E/F1 11/Times-Bold@0 SF(9.)72 524.8 Q
/F2 14/Times-Bold@0 SF(Security Considerations)5.5 E F0
(There are no security considerations introduced in this document.)100.8
541.4 Q
(General security considerations for TFRC are discussed in RFC 3448.)
100.8 558 Q(The security)5.5 E
(considerations for TFRC include the need to protect ag)100.8 571 Q
(ainst spoofed feedback, and the)-.055 E(need for protection mechanisms\
 to protect the congestion control mechanisms ag)100.8 584 Q(ainst)-.055
E(incorrect information from the recei)100.8 597 Q -.165(ve)-.275 G
-.605(r.).165 G(Security considerations for DCCP')100.8 613.6 Q 2.75(sC)
-.605 G(ongestion Control ID 3, TFRC Congestion Control,)-2.75 E
(are discussed in [CCID3].)100.8 626.6 Q(That document e)5.5 E(xtensi)
-.165 E -.165(ve)-.275 G(ly discussed the mechanisms the).165 E
(sender can use to v)100.8 639.6 Q
(erify the information sent by the recei)-.165 E -.165(ve)-.275 G -.605
(r.).165 G F1(10.)72 665.6 Q F2(IAN)5.5 E 3.5(AC)-.28 G(onsiderations)
-3.5 E F0(There are no IAN)100.8 682.2 Q 2.75(Ac)-.385 G
(onsiderations in this document.)-2.75 E(Flo)72 733 Q(yd/K)-.11 E
273.171(ohler Section)-.385 F 2.75(10. [P)2.75 F(age 18])-.165 E 0 Cg EP
%%Page: 19 19
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Times-Bold@0 SF(11.)72
85 Q/F2 14/Times-Bold@0 SF(Conclusions)5.5 E F0
(This document has speci\214ed TFRC-SP)100.8 101.6 Q 2.75(,aS)-1.221 G
(mall-P)-2.75 E(ack)-.165 E(et \(SP\) v)-.11 E
(ariant of TFRC, designed for)-.275 E(applications that send small pack)
100.8 114.6 Q(ets, with at most a hundred pack)-.11 E(ets per second, b)
-.11 E(ut that)-.22 E
(desire the same sending rate in bps as a TCP connection e)100.8 127.6 Q
(xperiencing the same pack)-.165 E(et drop)-.11 E(rate b)100.8 140.6 Q
(ut sending pack)-.22 E(ets of 1500 bytes.)-.11 E
(TFRC-SP competes reasonably f)5.5 E(airly with lar)-.11 E(ge-)-.198 E
(pack)100.8 153.6 Q(et TCP and TFRC \215o)-.11 E(ws in en)-.275 E
(vironments where lar)-.44 E(ge-pack)-.198 E(et \215o)-.11 E
(ws and small-pack)-.275 E(et)-.11 E<8d6f>100.8 166.6 Q(ws e)-.275 E
(xperience similar pack)-.165 E(et drop rates, b)-.11 E(ut recei)-.22 E
-.165(ve)-.275 G 2.75(sm).165 G(ore than its share of the)-2.75 E
(bandwidth in bps in en)100.8 179.6 Q(vironments where small pack)-.44 E
(ets are less lik)-.11 E(ely to be dropped or)-.11 E(mark)100.8 192.6 Q
(ed than are lar)-.11 E(ge pack)-.198 E 2.75(ets. As)-.11 F 2.75(ar)2.75
G(esult, TFRC-SP is e)-2.75 E(xperimental, and is not intended)-.165 E
(for widespread deplo)100.8 205.6 Q
(yment at this time in the global Internet.)-.11 E(In order to allo)
100.8 222.2 Q 2.75(we)-.275 G
(xperimentation with TFRC-SP in the Datagram Congestion Control)-2.915 E
(Protocol \(DCCP\), an e)100.8 235.2 Q
(xperimental Congestion Control IDenti\214er \(CCID\) will be used,)
-.165 E(based on TFRC-SP b)100.8 248.2 Q
(ut speci\214ed in a separate document.)-.22 E F1(12.)72 274.2 Q F2
(Thanks)5.5 E F0 1.76 -.88(We t)100.8 290.8 T(hank T).88 E
(om Phelan for discussions of TFRC-SP and for his paper e)-.88 E
(xploring the f)-.165 E(airness)-.11 E(between TCP and TFRC-SP \215o)
100.8 303.8 Q 2.75(ws. W)-.275 F 2.75(et)-.88 G
(hank Lars Eggert, Gorry F)-2.75 E(airhurst, Mark Handle)-.165 E -.715
(y,)-.165 G(Ladan Gharai, Colin Perkins, Pete Sholander)100.8 316.8 Q
2.75(,M)-.44 G(agnus W)-2.75 E(esterlund, and Joer)-.88 E 2.75(gW)-.198
G(idmer for)-3.19 E(feedback on earlier v)100.8 329.8 Q
(ersions of this draft.)-.165 E 1.76 -.88(We a)5.5 H
(lso thank the DCCP W).88 E(orking Group for)-.88 E
(feedback and discussions.)100.8 342.8 Q F1(A.)72 368.8 Q F2 -.35(Ap)5.5
G(pendix: Related W).35 E(ork on Small-P)-1.05 E(ack)-.14 E(et V)-.14 E
(ariants of TFRC)-1.288 E F0(Other proposals for v)100.8 385.4 Q
(ariants of TFRC for applications with v)-.275 E(ariable pack)-.275 E
(et sizes include)-.11 E([WBL04] and [V00]. [V00] proposed that TFRC sh\
ould be modi\214ed so that \215o)100.8 398.4 Q(ws are not)-.275 E
(penalized by sending smaller pack)100.8 411.4 Q 2.75(ets. In)-.11 F
(particular)2.75 E 2.75(,[)-.44 G(V00] proposes using the path MTU in)
-2.75 E(the TCP-friendly equation, instead of the actual pack)100.8
424.4 Q(et size used by TFRC, and counting)-.11 E(the pack)100.8 437.4 Q
(et drop rate by using an estimation algorithm that aggre)-.11 E -.055
(ga)-.165 G(tes both pack).055 E(et drops and)-.11 E(recei)100.8 450.4 Q
-.165(ve)-.275 G 2.75(dp).165 G(ack)-2.75 E(ets into MTU-sized units.)
-.11 E([WBL04] also ar)100.8 467 Q(gues that adapting TFRC for v)-.198 E
(ariable pack)-.275 E(et sizes by just using the pack)-.11 E(et)-.11 E
(size of a reference TCP \215o)100.8 480 Q 2.75(wi)-.275 G 2.75(nt)-2.75
G(he TFRC equation w)-2.75 E(ould not suf)-.11 E
(\214ce in the high-pack)-.275 E(et-)-.11 E(loss re)100.8 493 Q
(gime; such a modi\214ed TFRC w)-.165 E(ould ha)-.11 E .33 -.165(ve a s)
-.22 H(trong bias in f).165 E -.22(avo)-.11 G 2.75(ro).22 G 2.75(fs)
-2.75 G(maller pack)-2.75 E(ets,)-.11 E(because multiple lost pack)100.8
506 Q(ets in a single round-trip time w)-.11 E(ould be aggre)-.11 E
-.055(ga)-.165 G(ted into a single).055 E(pack)100.8 519 Q(et loss.)-.11
E([WBL04] proposes modifying the loss measurement process to account fo\
r the)5.5 E(bias in f)100.8 532 Q -.22(avo)-.11 G 2.75(ro).22 G 2.75(fs)
-2.75 G(maller pack)-2.75 E(ets.)-.11 E(The TFRC-SP v)100.8 548.6 Q
(ariant of TFRC proposed in our document dif)-.275 E
(fers from [WBL04] in)-.275 E(restricting its attention to \215o)100.8
561.6 Q(ws that send at most 100 pack)-.275 E(ets per second.)-.11 E
(The TFRC-SP)5.5 E -.275(va)100.8 574.6 S
(riant proposed in our document also dif).275 E(fers from the stra)-.275
E 2.75(wp)-.165 G(roposal discussed in)-2.75 E([WBL04] in that the allo)
100.8 587.6 Q(wed bandwidth includes the bandwidth used by pack)-.275 E
(et headers.)-.11 E([WBL04] discusses four methods for modifying the lo\
ss measurement process,)100.8 604.2 Q("unbiasing", "virtual pack)100.8
617.2 Q(ets", "random sampling", and "Loss Insensiti)-.11 E .33 -.165
(ve P)-.275 H(eriod \(LIP\)).165 E 2.75(scaling". [WBL04])100.8 630.2 R
(\214nds only the second and third methods suf)2.75 E(\214ciently rob)
-.275 E(ust when the)-.22 E(netw)100.8 643.2 Q(ork drops pack)-.11 E
(ets independently of pack)-.11 E(et size.)-.11 E(The)5.5 E 2.75<798c>
-.165 G(nd only the second method)-2.75 E(suf)100.8 656.2 Q
(\214ciently rob)-.275 E(ust when the netw)-.22 E(ork is more lik)-.11 E
(ely to drop lar)-.11 E(ge pack)-.198 E(ets than small)-.11 E(pack)100.8
669.2 Q 2.75(ets. Our)-.11 F(method for calculating the loss e)2.75 E
-.165(ve)-.275 G(nt rate is some).165 E(what similar to the random)-.275
E(sampling method proposed in [WBL04], e)100.8 682.2 Q
(xcept that randomization is not used; instead of)-.165 E
(randomization, the e)100.8 695.2 Q(xact pack)-.165 E
(et loss rate is computed for short loss interv)-.11 E(als, and the)
-.275 E(Flo)72 733 Q(yd/K)-.11 E 276.229(ohler Section)-.385 F 2.75
(A. [P)2.75 F(age 19])-.165 E 0 Cg EP
%%Page: 20 20
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(standard loss e)100.8 85 Q
-.165(ve)-.275 G(nt rate calculation is used for longer loss interv).165
E 2.75(als. [WBL04])-.275 F(includes)2.75 E(simulations with a Bernoull\
i loss model, a Bernoulli loss model with a drop rate v)100.8 98 Q
(arying)-.275 E -.165(ove)100.8 111 S 2.75(rt).165 G(ime, and a Gilbert\
 loss model, as well as more realistic simulations with a range of)-2.75
E(TCP and TFRC \215o)100.8 124 Q(ws.)-.275 E
([WBL04] produces both a byte-mode and a pack)100.8 140.6 Q(et-mode v)
-.11 E(ariant of the TFRC transport)-.275 E(protocol, for connections o)
100.8 153.6 Q -.165(ve)-.165 G 2.75(rp).165 G(aths with per)-2.75 E
(-byte and per)-.22 E(-pack)-.22 E(et dropping respecti)-.11 E -.165(ve)
-.275 G(ly).165 E(.)-.715 E 1.76 -.88(We w)100.8 166.6 T(ould ar).77 E
(gue that in the absence of transport-le)-.198 E -.165(ve)-.275 G 2.75
(lm).165 G(echanisms for determining whether)-2.75 E(the pack)100.8
179.6 Q(et dropping in the netw)-.11 E(ork is per)-.11 E(-pack)-.22 E
(et, per)-.11 E(-byte, or some)-.22 E(where in between, a)-.275 E
(single TFRC implementation is needed, independently of the pack)100.8
192.6 Q(et-dropping beha)-.11 E(viors of)-.22 E
(the routers along the path.)100.8 205.6 Q(It w)5.5 E
(ould of course be preferable to ha)-.11 E .33 -.165(ve a m)-.22 H
(echanism that gi).165 E -.165(ve)-.275 G(s).165 E
(roughly acceptable beha)100.8 218.6 Q(vior)-.22 E 2.75(,t)-.44 G 2.75
(ot)-2.75 G(he connection and to the netw)-2.75 E
(ork as a whole, on paths)-.11 E(with both per)100.8 231.6 Q
(-byte and per)-.22 E(-pack)-.22 E
(et dropping \(and on paths with multiple congested routers,)-.11 E
(some with per)100.8 244.6 Q(-byte dropping mechanisms, some with per)
-.22 E(-pack)-.22 E(et dropping mechanisms,)-.11 E
(and some with dropping mechanisms that lie some)100.8 257.6 Q
(where between per)-.275 E(-byte and per)-.22 E(-pack)-.22 E(et\).)-.11
E(An important contrib)100.8 274.2 Q(ution w)-.22 E(ould be to in)-.11 E
-.165(ve)-.44 G(stig).165 E(ate the range of beha)-.055 E
(viors actually present)-.22 E(in today')100.8 287.2 Q 2.75(sn)-.605 G
(etw)-2.75 E(orks, in terms of pack)-.11 E
(et-dropping as a function of pack)-.11 E(et size.)-.11 E/F1 11
/Times-Bold@0 SF(B.)72 313.2 Q/F2 14/Times-Bold@0 SF(Simulation Results)
5.5 E F0(This appendix reports on the simulation results outlined in Se\
ction 7 . TFRC-SP has been)100.8 329.8 Q(added to the NS simulator)100.8
342.8 Q 2.75(,a)-.44 G(nd is illustrated in the v)-2.75 E
(alidation test "./test-all-friendly" in the)-.275 E
(directory tcl/tests.)100.8 355.8 Q
(The simulation scripts and graphs for the simulations in this document)
5.5 E(are a)100.8 368.8 Q -.275(va)-.22 G(ilable at [V).275 E(OIPSIMS].)
-.44 E F1(B.1.)72 394.8 Q/F3 13/Times-Bold@0 SF
(Simulations with Con\214gur)5.5 E(ed P)-.234 E(ack)-.13 E(et Dr)-.13 E
(op Rates)-.234 E F0(In this section we describe simulation results fro\
m simulations comparing the throughput)100.8 411.4 Q(of standard \(SA)
100.8 424.4 Q(CK\) TCP \215o)-.44 E(ws, TCP \215o)-.275 E
(ws with timestamps and ECN, TFRC-SP \215o)-.275 E(ws,)-.275 E
(and standard TFRC \(Stnd TFRC\) \215o)100.8 437.4 Q 2.75(ws. In)-.275 F
(these simulations we con\214gure the router to)2.75 E
(randomly drop or mark pack)100.8 450.4 Q
(ets at a speci\214ed rate, independently of the pack)-.11 E(et size.)
-.11 E -.165(Fo)5.5 G(r).165 E(each speci\214ed pack)100.8 463.4 Q
(et drop rate, we gi)-.11 E .33 -.165(ve a \215)-.275 H -.275(ow).165 G
1.21 -.605('s a).275 H -.165(ve).385 G(rage sending rate in Kbps o).165
E -.165(ve)-.165 G 2.75(rt).165 G(he)-2.75 E
(second half of a 100-second simulation, a)100.8 476.4 Q -.165(ve)-.22 G
(raged o).165 E -.165(ve)-.165 G 2.75(rt).165 G(en \215o)-2.75 E(ws.)
-.275 E(Flo)72 733 Q(yd/K)-.11 E 268.584(ohler Section)-.385 F 2.75
(B.1. [P)2.75 F(age 20])-.165 E 0 Cg EP
%%Page: 21 21
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 39.6
(Packet Send)180 85 R(Rates \(Kbps, 1460B seg\))6.6 E 33
(DropRate TCP ECN)180 98 R 33(TCP TFRC)6.6 F 19.8(-------- --------)180
111 R 13.2(-------- --------)19.8 F 19.8(0.001 2020.85 1904.61)199.8 124
R(982.09)33 E 26.4(0.005 811.10 792.11 878.08)199.8 137 R 33
(0.01 515.45)199.8 150 R 26.4(533.19 598.90)33 F 33(0.02 362.93)199.8
163 R 26.4(382.67 431.41)33 F 33(0.04 250.06)199.8 176 R 26.4
(252.64 284.82)33 F 33(0.05 204.48)199.8 189 R 26.4(218.16 268.51)33 F
39.6(0.1 143.30)199.8 202 R 26.4(148.41 146.03)33 F 46.2(0.2 78.65)199.8
215 R 26.4(93.23* 55.14)39.6 F 46.2(0.3 26.26)199.8 228 R 26.4
(59.65* 32.87)39.6 F 52.8(0.4 9.87)199.8 241 R 26.4(47.79* 25.45)39.6 F
52.8(0.5 3.53)199.8 254 R 26.4(32.01* 18.52)39.6 F 6.6(*E)133.8 280 S
(CN scenarios marked with an asterisk are not realistic,)-6.6 E
(as routers are not recommended to mark packets when packet)147 293 Q
(drop/mark rates are 20% or higher.)147 306 Q
(Table 7: Send Rate vs. Packet Drop Rate I:)180 332 Q
(1460B TFRC Segments)272.4 345 Q
(\(1.184 Kbps Maximum TFRC Data Sending Rate\))193.2 358 Q F0 -.88(Ta)
100.8 377 S(ble 7 sho).88 E
(ws the sending rate for a TCP and a standard TFRC \215o)-.275 E 2.75
(wf)-.275 G(or a range of)-2.75 E(con\214gured pack)100.8 390 Q
(et drop rates, when both \215o)-.11 E(ws ha)-.275 E .33 -.165(ve 1)-.22
H(460-byte data se).165 E(gments, in order to)-.165 E
(illustrate the relati)100.8 403 Q .33 -.165(ve f)-.275 H
(airness of TCP and TFRC when both \215o).055 E(ws use the same pack)
-.275 E(et size.)-.11 E -.165(Fo)100.8 416 S 2.75(re).165 G
(xample, a pack)-2.915 E
(et drop rate of 0.1 means that 10% of the TCP and TFRC pack)-.11 E
(ets are)-.11 E 2.75(dropped. The)100.8 429 R(TFRC \215o)2.75 E 2.75(wi)
-.275 G 2.75(sc)-2.75 G(on\214gured to send at most 100 pack)-2.75 E
(ets per second.)-.11 E(There is)5.5 E(good relati)100.8 442 Q .33 -.165
(ve f)-.275 H(airness until the pack).055 E
(et drop percentages reach 40 and 50%, when the TFRC)-.11 E<8d6f>100.8
455 Q 2.75(wr)-.275 G(ecei)-2.75 E -.165(ve)-.275 G 2.75(st).165 G
(hree to \214v)-2.75 E 2.75(et)-.165 G
(imes more bandwidth than the standard TCP \215o)-2.75 E 4.18 -.715
(w. W)-.275 H 2.75(en)-.165 G(ote that)-2.75 E(an ECN TCP \215o)100.8
468 Q 2.75(ww)-.275 G(ould recei)-2.86 E .33 -.165(ve a h)-.275 H
(igher throughput than the TFRC \215o).165 E 2.75(wa)-.275 G 2.75(tt)
-2.75 G(hese high)-2.75 E(pack)100.8 481 Q
(et drop rate, if ECN-marking w)-.11 E(as still being instead of pack)
-.11 E 2.75(et-dropping. Ho)-.11 F(we)-.275 E -.165(ve)-.275 G .88 -.44
(r, w).165 H(e).44 E(don')100.8 494 Q 2.75(tu)-.198 G
(se the ECN TCP sending rate in these high-pack)-2.75 E
(et-drop scenarios as the tar)-.11 E(get)-.198 E(sending rate for TFRC,\
 as routers are advised to drop rather than mark pack)100.8 507 Q
(ets during high)-.11 E(le)100.8 520 Q -.165(ve)-.275 G
(ls of congestion [RFC3168] \(Section 7\).).165 E
(In addition, there is lik)5.5 E(ely to be b)-.11 E(uf)-.22 E(fer)-.275
E -.165(ove)100.8 533 S(r\215o).165 E 2.75(wi)-.275 G 2.75(ns)-2.75 G
(cenarios with such high pack)-2.75 E
(et dropping/marking rates, forcing routers to drop)-.11 E(pack)100.8
546 Q(ets instead of ECN-marking them.)-.11 E(Flo)72 733 Q(yd/K)-.11 E
268.584(ohler Section)-.385 F 2.75(B.1. [P)2.75 F(age 21])-.165 E 0 Cg
EP
%%Page: 22 22
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<------S)206.4 85 S(end Rates \(Kbps\) - - - - - >)-6.6 E 39.6
(Packet TCP ECN)147 98 R 26.4(TCP TFRC-SP)6.6 F(Stnd TFRC)19.8 E 6.6
(DropRate \(1460B)140.4 111 R(seg\) \(1460B seg\))6.6 E(\(14B seg\))13.2
E(\(14B seg\))13.2 E 6.6(-------- -----------)140.4 124 R 6.6
(----------- --------- ---------)6.6 F 19.8(0.001 1787.54)160.2 137 R 33
(1993.03 17.71 17.69)33 F 26.4(0.005 785.11)160.2 150 R 33
(823.75 18.11 17.69)39.6 F 33(0.01 533.38 529.01 17.69 17.80)160.2 163 R
33(0.02 317.16 399.62 17.69 13.41)160.2 176 R 33
(0.04 245.42 260.57 17.69)160.2 189 R(8.84)46.2 E 33
(0.05 216.38 223.75 17.69)160.2 202 R(7.63)46.2 E 39.6(0.1 142.75)160.2
215 R 33(138.36 17.69)39.6 F(4.29)46.2 E 46.2(0.2 58.61)160.2 228 R 26.4
(91.54* 17.80)46.2 F(1.94)46.2 E 46.2(0.3 21.62)160.2 241 R 26.4
(63.96* 10.26)46.2 F(1.00)46.2 E 46.2(0.4 10.51)160.2 254 R 33
(41.74* 4.78)46.2 F(0.77)46.2 E 52.8(0.5 1.92)160.2 267 R 33
(19.03* 2.41)46.2 F(0.56)46.2 E 6.6(*E)133.8 293 S
(CN scenarios marked with an asterisk are not realistic,)-6.6 E
(as routers are not recommended to mark packets when packet)147 306 Q
(drop/mark rates are 20% or higher.)147 319 Q
(Table 8: Send Rate vs. Packet Drop Rate II:)180 345 Q
(14B TFRC Segments)279 358 Q
(\(5.6 Kbps Maximum TFRC Data Sending Rate\))199.8 371 Q F0 -.88(Ta)
100.8 393.6 S(ble 8 sho).88 E
(ws the results of simulations where each TFRC-SP \215o)-.275 E 2.75(wh)
-.275 G(as a maximum data)-2.75 E
(sending rate of 5.6 Kbps, with 14-byte data pack)100.8 406.6 Q
(ets and a 32-byte pack)-.11 E(et header for DCCP)-.11 E(and IP)100.8
419.6 Q 5.5(.E)-1.221 G(ach TCP \215o)-5.5 E 2.75(wh)-.275 G(as a recei)
-2.75 E .33 -.165(ve w)-.275 H(indo).165 E 2.75(wo)-.275 G 2.75(f1)-2.75
G(00 pack)-2.75 E(ets and a data pack)-.11 E(et size of 1460)-.11 E
(bytes, with a 40-byte pack)100.8 432.6 Q(et header for TCP and IP)-.11
E 5.5(.T)-1.221 G(he TCP \215o)-5.5 E 2.75(wu)-.275 G(ses SA)-2.75 E
(CK TCP with)-.44 E(Limited T)100.8 445.6 Q(ransmit, b)-.385 E
(ut without timestamps or ECN.)-.22 E(Each \215o)5.5 E 2.75(wh)-.275 G
(as a round-trip time of 240)-2.75 E
(ms in the absence of queueing delay)100.8 458.6 Q(.)-.715 E
(The TFRC sending rate in T)100.8 475.2 Q
(able 8 is the sending rate for the 14-byte data pack)-.88 E
(et with the)-.11 E(32-byte pack)100.8 488.2 Q(et header)-.11 E 5.5(.T)
-.605 G(hus, only 30% of the TFRC sending rate is for data, and with a)
-5.5 E(pack)100.8 501.2 Q
(et drop rate of p, only a fraction 1-p of that data mak)-.11 E
(es it to the recei)-.11 E -.165(ve)-.275 G 3.96 -.605(r. T).165 H
(hus, the).605 E(TFRC data recei)100.8 514.2 Q .33 -.165(ve r)-.275 H
(ate can be considerably less than the TFRC sending rate in the table.)
.165 E(Because TCP uses lar)100.8 527.2 Q(ge pack)-.198 E
(ets, 97% of the TCP sending rate is for data, and the same)-.11 E
(fraction 1-p of that data mak)100.8 540.2 Q(es it to the recei)-.11 E
-.165(ve)-.275 G -.605(r.).165 G -.88(Ta)100.8 556.8 S(ble 8 sho).88 E(\
ws that for the 5.6 Kbps data stream with TFRC, Standard TFRC \(Stnd TF\
RC\))-.275 E(gi)100.8 569.8 Q -.165(ve)-.275 G 2.75(sav).165 G
(ery poor sending rate in bps, relati)-2.915 E .33 -.165(ve t)-.275 H
2.75(ot).165 G(he sending rate for the lar)-2.75 E(ge-pack)-.198 E
(et TCP)-.11 E 2.75(connection. In)100.8 582.8 R
(contrast, the sending rate for the TFRC-SP \215o)2.75 E 2.75(wi)-.275 G
2.75(sr)-2.75 G(elati)-2.75 E -.165(ve)-.275 G(ly close to the).165 E
(desired goal of f)100.8 595.8 Q(airness in bps with TCP)-.11 E(.)-1.221
E -.88(Ta)100.8 612.4 S(ble 8 sho).88 E(ws that with TFRC-SP)-.275 E
2.75(,t)-1.221 G(he 5.6 Kbps data stream doesn')-2.75 E 2.75(tr)-.198 G
(educe its sending rate)-2.75 E(until pack)100.8 625.4 Q
(et drop rates greater than 20%, as desired.)-.11 E -.44(Wi)5.5 G
(th pack).44 E(et drop rates of 30-40%, the)-.11 E
(sending rate for the TFRC-SP \215o)100.8 638.4 Q 2.75(wi)-.275 G 2.75
(ss)-2.75 G(ome)-2.75 E(what less than that of the a)-.275 E -.165(ve)
-.22 G(rage lar).165 E(ge-pack)-.198 E(et)-.11 E(TCP \215o)100.8 651.4 Q
1.43 -.715(w, w)-.275 H(hile for pack).715 E
(et drop rates of 50% the sending rate for the TFRC-SP \215o)-.11 E 2.75
(wi)-.275 G(s)-2.75 E(some)100.8 664.4 Q
(what greater than that of the a)-.275 E -.165(ve)-.22 G(rage lar).165 E
(ge-pack)-.198 E(et TCP \215o)-.11 E -.715(w.)-.275 G(Flo)72 733 Q(yd/K)
-.11 E 268.584(ohler Section)-.385 F 2.75(B.1. [P)2.75 F(age 22])-.165 E
0 Cg EP
%%Page: 23 23
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<------S)206.4 85 S(end Rates \(Kbps\) - - - - - >)-6.6 E 39.6
(Packet TCP ECN)147 98 R 19.8(TCP TFRC-SP)6.6 F(Stnd TFRC)19.8 E 6.6
(DropRate \(1460B)140.4 111 R
(seg\) \(1460B seg\) \(200B seg\) \(200B seg\))6.6 E 6.6
(-------- -----------)140.4 124 R(----------- ---------- ----------)6.6
E 19.8(0.001 1908.98)160.2 137 R 26.4(1869.24 183.45 178.35)33 F 26.4
(0.005 854.69)160.2 150 R 26.4(835.10 185.06 138.06)39.6 F 33
(0.01 564.10 531.10)160.2 163 R 33(185.33 92.43)33 F 33
(0.02 365.38 369.10)160.2 176 R 33(185.57 62.18)33 F 33
(0.04 220.80 257.81)160.2 189 R 33(185.14 45.43)33 F 33
(0.05 208.97 219.41)160.2 202 R 33(180.08 39.44)33 F 39.6(0.1 141.67)
160.2 215 R 26.4(143.88 127.33)39.6 F(21.96)39.6 E 46.2(0.2 62.66)160.2
228 R 26.4(91.87* 54.66)46.2 F(9.40)46.2 E 46.2(0.3 16.63)160.2 241 R
26.4(65.52* 24.50)46.2 F(4.73)46.2 E 52.8(0.4 6.62)160.2 254 R 26.4
(42.00* 13.47)46.2 F(3.35)46.2 E 52.8(0.5 1.01)160.2 267 R 26.4
(21.34* 10.51)46.2 F(2.92)46.2 E 6.6(*E)133.8 293 S
(CN scenarios marked with an asterisk are not realistic,)-6.6 E
(as routers are not recommended to mark packets when packet)147 306 Q
(drop/mark rates are 20% or higher.)147 319 Q
(Table 9: Sending Rate vs. Packet Drop Rate III:)180 345 Q
(200B TFRC Segments)265.8 358 Q
(\(160 Kbps Maximum TFRC Data Sending Rate\))186.6 371 Q F0 -.88(Ta)
100.8 390 S(ble 9 sho).88 E(ws results with con\214gured pack)-.275 E
(et drop rates when the TFRC \215o)-.11 E 2.75(wu)-.275 G(ses 200-byte)
-2.75 E(data pack)100.8 403 Q
(ets, with a maximum data sending rate of 160 Kbps.)-.11 E(As in T)5.5 E
(able 8, the)-.88 E(performance of Standard TFRC is quite poor)100.8 416
Q 2.75(,w)-.44 G(hile the performance of TFRC-SP is)-2.75 E
(essentially as desired for pack)100.8 429 Q(et drop rates up to 30%.)
-.11 E(Ag)5.5 E(ain as e)-.055 E(xpected, with pack)-.165 E(et drop)-.11
E(rates of 40-50% the TFRC-SP sending rate is some)100.8 442 Q
(what higher than desired.)-.275 E -.165(Fo)100.8 458.6 S 2.75(rt).165 G
(hese simulations, the sending rate of a TCP connection using timestamp\
s is similar to)-2.75 E(the sending rate sho)100.8 471.6 Q
(wn for a standard TCP connection without timestamps.)-.275 E/F2 11
/Times-Bold@0 SF(B.2.)72 497.6 Q/F3 13/Times-Bold@0 SF
(Simulations with Con\214gur)5.5 E(ed Byte Dr)-.234 E(op Rates)-.234 E
F0(In this section we e)100.8 514.2 Q(xplore simulations where the rout\
er is con\214gured to drop or mark each)-.165 E
(*byte* at a speci\214ed rate.)100.8 527.2 Q
(When a byte is chosen to be dropped \(or mark)5.5 E(ed\), the entire)
-.11 E(pack)100.8 540.2 Q(et containing that byte is dropped \(or mark)
-.11 E(ed\).)-.11 E(Flo)72 733 Q(yd/K)-.11 E 268.584(ohler Section)-.385
F 2.75(B.2. [P)2.75 F(age 23])-.165 E 0 Cg EP
%%Page: 24 24
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<-----S)259.2 85 S(end Rates \(Kbps\) - - - - - >)-6.6 E 39.6(Byte TCP)
133.8 98 R 19.8(TFRC-SP Stnd)178.2 F(TFRC)6.6 E 13.2(DropRate SegSize)
120.6 111 R 33(TCP ECN)33 F 19.8(TCP \(14B)6.6 F 6.6(seg\) \(14B)6.6 F
(seg\))6.6 E 13.2(-------- ------- -------- -------- ---------)120.6 124
R(---------)13.2 E 26.4(0.00001 1460 423.02 431.26)127.2 137 R 33
(17.69 17.69)39.6 F 33(0.0001 1460)127.2 150 R 26.4(117.41 114.34)33 F
33(17.69 17.69)39.6 F 46.2(0.001 128)127.2 163 R 33(10.78 11.50 17.69)
39.6 F(8.37)46.2 E 52.8(0.005 14)127.2 176 R 39.6(1.75 2.89)46.2 F 39.6
(18.39 1.91)39.6 F 39.6(0.010 1460 0.31 0.26 7.07 0.84)127.2 189 R 39.6
(0.020 1460 0.29 0.26 1.61 0.43)127.2 202 R 39.6(0.040 1460 0.12 0.26*)
127.2 215 R 39.6(0.17 0.12)39.6 F 39.6(0.050 1460 0.15 0.26*)127.2 228 R
39.6(0.08 0.06)39.6 F 6.6(*E)133.8 254 S
(CN scenarios marked with an asterisk are not realistic,)-6.6 E
(as routers are not recommended to mark packets when packet)147 267 Q
(drop/mark rates are 20% or higher.)147 280 Q
(TFRC's maximum data sending rate is 5.6 Kbps.)147 306 Q
(Table 10: Sending Rate vs. Byte Drop Rate)153.6 332 Q F0 -.88(Ta)100.8
354.6 S(ble 10 sho).88 E(ws the TCP and TFRC send rates for v)-.275 E
(arious byte drop rates.)-.275 E -.165(Fo)5.5 G 2.75(re).165 G(ach byte)
-2.75 E(drop rate, T)100.8 367.6 Q(able 10 sho)-.88 E
(ws the TCP sending rate, with and without ECN, for the TCP)-.275 E(se)
100.8 380.6 Q(gment size that gi)-.165 E -.165(ve)-.275 G 2.75(st).165 G
(he highest TCP sending rate.)-2.75 E(Simulations were run with TCP)5.5
E(se)100.8 393.6 Q(gments of 14, 128, 256, 512, and 1460 bytes.)-.165 E
(Thus, for a byte drop rate of 0.00001, the)5.5 E(table sho)100.8 406.6
Q(ws the TCP sending rate with 1460-byte data se)-.275 E(gments, b)-.165
E(ut with a byte drop rate)-.22 E(of 0.001, the table sho)100.8 419.6 Q
(ws the TCP sending rate with 128-byte data se)-.275 E 2.75(gments. F)
-.165 F(or each byte)-.165 E(drop rate, T)100.8 432.6 Q
(able 10 also sho)-.88 E
(ws the TFRC-SP and that Standard TFRC sending rates.)-.275 E -.44(Wi)
5.5 G(th).44 E(con\214gured byte drop rates, TFRC-SP gi)100.8 445.6 Q
-.165(ve)-.275 G 2.75(sa).165 G 2.75(nu)-2.75 G(nf)-2.75 E(air adv)-.11
E(antage to the TFRC-SP \215o)-.275 E -.715(w,)-.275 G
(while Standard TFRC gi)100.8 458.6 Q -.165(ve)-.275 G 2.75(se).165 G
(ssentially the desired performance.)-2.75 E F1(TCP Pkt)232.8 477.6 Q
(TFRC Pkt)33 E 26.4(Byte DropRate)173.4 490.6 R 39.6(DropRate TCP/TFRC)
26.4 F 6.6(DropRate \(1460B)160.2 503.6 R 6.6(seg\) \(14B)6.6 F 13.2
(seg\) Pkt)6.6 F(Drop Ratio)6.6 E 6.6(-------- ----------- ---------)
160.2 516.6 R(--------------)19.8 E 26.4(0.00001 0.015)166.8 529.6 R
46.2(0.0006 26.59)46.2 F 33(0.0001 0.13)166.8 542.6 R 46.2(0.0056 24.94)
52.8 F 39.6(0.001 0.77)166.8 555.6 R 52.8(0.054 14.26)52.8 F 39.6
(0.005 0.99)166.8 568.6 R 66(0.24 4.08)52.8 F 46.2(0.01 1.00 0.43)166.8
581.6 R(2.32)72.6 E 46.2(0.05 1.00 0.94)166.8 594.6 R(1.05)72.6 E
(Table 11: Packet Drop Rate Ratio vs. Byte Drop Rate)153.6 620.6 Q F0
-.88(Ta)100.8 643.2 S(ble 11 con).88 E -.165(ve)-.44 G
(rts the byte drop rate p to pack).165 E
(et drop rates for the TCP and TFRC pack)-.11 E(ets,)-.11 E
(where the pack)100.8 656.2 Q(et drop rate for an N-byte pack)-.11 E
(et is 1-\(1-p\)^N.)-.11 E(Thus, a byte drop rate of)5.5 E
(0.001, with each byte being dropped with probability 0.001, con)100.8
669.2 Q -.165(ve)-.44 G(rts to a pack).165 E(et drop rate)-.11 E
(of 0.77, or 77%, for the 1500-byte TCP pack)100.8 682.2 Q
(ets, and a pack)-.11 E(et drop rate of 0.054, or 5.4%,)-.11 E
(for the 56-byte TFRC pack)100.8 695.2 Q(ets.)-.11 E(Flo)72 733 Q(yd/K)
-.11 E 268.584(ohler Section)-.385 F 2.75(B.2. [P)2.75 F(age 24])-.165 E
0 Cg EP
%%Page: 25 25
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(The right column of T)100.8 85
Q(able 11 sho)-.88 E(ws the ratio between the TCP pack)-.275 E
(et drop rate and the)-.11 E(TFRC pack)100.8 98 Q(et drop rate.)-.11 E
-.165(Fo)5.5 G 2.75(rl).165 G .55 -.275(ow b)-2.75 H
(yte drop rates, this ratio is close to 26.8, the ratio).275 E
(between the TCP and TFRC pack)100.8 111 Q(et sizes.)-.11 E -.165(Fo)5.5
G 2.75(rh).165 G(igh byte drop rates, where e)-2.75 E -.165(ve)-.275 G
2.75(nm).165 G(ost small)-2.75 E(TFRC pack)100.8 124 Q(ets are lik)-.11
E(ely to be dropped, this drop ratio approaches 1.)-.11 E(As T)5.5 E
(able 10 sho)-.88 E(ws,)-.275 E(with byte drop rates, the Standard TFRC\
 sending rate is close to optimal, competing f)100.8 137 Q(airly)-.11 E
(with a TCP connection using the optimal pack)100.8 150 Q
(et size within the allo)-.11 E(wed range.)-.275 E(In contrast,)5.5 E(t\
he TFRC-SP connection gets more than its share of the bandwidth, though\
 it does reduce)100.8 163 Q(its sending rate for a byte drop rate of 0.\
01 or more \(corresponding to a TFRC-SP *pack)100.8 176 Q(et*)-.11 E
(drop rate of 0.43.)100.8 189 Q -.88(Ta)100.8 205.6 S
(ble 10 essentially sho).88 E(ws three separate re)-.275 E 2.75
(gions. In)-.165 F(the lo)2.75 E(w-congestion re)-.275 E
(gion, with byte)-.165 E(drop rates less than 0.0001, the TFRC-SP conne\
ction is sending at its maximum sending)100.8 218.6 Q 2.75(rate. In)
100.8 231.6 R(this re)2.75 E
(gion the optimal TCP connection is the one with 1460-byte se)-.165 E
(gments, and)-.165 E(the TCP sending rate goes as 1/sqrt\(p\), for pack)
100.8 244.6 Q(et drop rate p.)-.11 E(This lo)5.5 E(w-congestion re)-.275
E(gion)-.165 E(holds for pack)100.8 257.6 Q(et drop rates up to 10-15%.)
-.11 E(In the middle re)100.8 274.2 Q(gion of T)-.165 E
(able 10, with byte drop rates from 0.0001 to 0.005, the optimal)-.88 E
(TCP se)100.8 287.2 Q(gment size is a function of the byte drop rate.)
-.165 E(In particular)5.5 E 2.75(,t)-.44 G(he optimal TCP)-2.75 E(se)
100.8 300.2 Q(gment size seems to be the one that k)-.165 E
(eeps the pack)-.11 E(et drop rate at most 15%, k)-.11 E(eeping the)-.11
E(TCP connection in the re)100.8 313.2 Q
(gime controlled by a 1/sqrt\(p\) sending rate, for pack)-.165 E
(et drop rate p.)-.11 E -.165(Fo)100.8 326.2 S 2.75(raT).165 G(CP pack)
-2.75 E(et size of S bytes \(including headers\), and a *byte* drop rat\
e of B, the)-.11 E(pack)100.8 339.2 Q(et drop rate is roughly B*S.)-.11
E -.165(Fo)5.5 G 2.75(rag).165 G -2.365 -.275(iv e)-2.75 H 2.75(nb).275
G(yte drop rate in this re)-2.75 E(gime, if the optimal)-.165 E
(TCP performance occurs with a pack)100.8 352.2 Q(et size chosen to gi)
-.11 E .33 -.165(ve a p)-.275 H(ack).165 E(et drop rate of at most)-.11
E(15%, k)100.8 365.2 Q(eeping the TCP connection out of the re)-.11 E
(gime of e)-.165 E(xponential back)-.165 E(of)-.11 E(fs of the)-.275 E
(retransmit timer)100.8 378.2 Q 2.75(,t)-.44 G
(hen this requires B*S = 0.15, or S = 0.15/B.)-2.75 E
(In the high-congestion re)100.8 394.8 Q(gime of T)-.165 E
(able 10, with high congestion and with byte drop rates)-.88 E
(of 0.01 and higher)100.8 407.8 Q 2.75(,t)-.44 G
(he TCP performance is dominated by the e)-2.75 E(xponential back)-.165
E(of)-.11 E 2.75(fo)-.275 G 2.75(ft)-2.75 G(he)-2.75 E
(retransmit timer re)100.8 420.8 Q -.055(ga)-.165 G(rdless of the se)
.055 E(gment size.)-.165 E(Ev)5.5 E(en a 40-byte pack)-.165 E
(et with a zero-byte)-.11 E(data se)100.8 433.8 Q(gment w)-.165 E
(ould ha)-.11 E .33 -.165(ve a p)-.22 H(ack).165 E
(et drop rate of at least 33%.)-.11 E(In this re)5.5 E
(gime, the optimal)-.165 E(TCP *sending* rate comes with a lar)100.8
446.8 Q(ge, 1460-byte data se)-.198 E(gment, b)-.165 E
(ut the TCP sending rate)-.22 E(is lo)100.8 459.8 Q 2.75(ww)-.275 G
(ith an)-2.75 E 2.75(ys)-.165 G -.165(eg)-2.75 G
(ment size, considerably less than one pack).165 E
(et per round-trip time.)-.11 E -.88(We)5.5 G(note that in this re)100.8
472.8 Q(gime, while a lar)-.165 E(ger pack)-.198 E(et gi)-.11 E -.165
(ve)-.275 G 2.75(sah).165 G(igher TCP *sending* rate, a smaller)-2.75 E
(pack)100.8 485.8 Q(et gi)-.11 E -.165(ve)-.275 G 2.75(sab).165 G
(etter *goodput* rate.)-2.75 E(In general, T)100.8 502.4 Q
(ables 8 and 9 sho)-.88 E 2.75(wg)-.275 G
(ood performance for TFRC-SP in en)-2.75 E(vironments with)-.44 E
(stable pack)100.8 515.4 Q
(et drop rates, where the decision to drop a pack)-.11 E
(et is independent of the pack)-.11 E(et)-.11 E 2.75(size. Ho)100.8
528.4 R(we)-.275 E -.165(ve)-.275 G .88 -.44(r, i).165 H 2.75(ns).44 G
(ome en)-2.75 E(vironments the pack)-.44 E(et size might af)-.11 E
(fect the lik)-.275 E(elihood that a)-.11 E(pack)100.8 541.4 Q
(et is dropped.)-.11 E -.165(Fo)5.5 G 2.75(re).165 G(xample, with hea)
-2.915 E(vy congestion and a Drop T)-.22 E(ail queue with a \214x)-.88 E
(ed)-.165 E(number of bytes rather than a \214x)100.8 554.4 Q
(ed number of pack)-.165 E(et-sized b)-.11 E(uf)-.22 E(fers, small pack)
-.275 E(ets might be)-.11 E(more lik)100.8 567.4 Q(ely than lar)-.11 E
(ge pack)-.198 E(ets to \214nd room at the end of an almost-full queue.)
-.11 E(As a)8.25 E(further complication, in a scenario with Acti)100.8
580.4 Q .33 -.165(ve Q)-.275 H(ueue Management, the A).165 E
(QM mechanism)-.605 E(could either be in pack)100.8 593.4 Q
(et mode, dropping each pack)-.11 E(et with equal probability)-.11 E
2.75(,o)-.715 G 2.75(ri)-2.75 G 2.75(nb)-2.75 G(yte)-2.75 E
(mode, dropping each byte with equal probability)100.8 606.4 Q 5.5(.S)
-.715 G(ections B.3 and B.4 sho)-5.5 E 2.75(ws)-.275 G(imulations)-2.75
E(with pack)100.8 619.4 Q(ets dropped at Drop T)-.11 E(ail or A)-.88 E
(QM queues, rather that from a probabilistic process.)-.605 E/F1 11
/Times-Bold@0 SF(B.3.)72 645.4 Q/F2 13/Times-Bold@0 SF -.13(Pa)5.5 G(ck)
.13 E(et Dr)-.13 E(opping Beha)-.234 E(vior at Routers with Dr)-.325 E
(op-T)-.234 E(ail Queues)-1.196 E F0
(One of the problems with comparing the throughput of tw)100.8 662 Q
2.75<6f8d>-.11 G -.275(ow)-2.75 G 2.75(su).275 G(sing dif)-2.75 E
(ferent pack)-.275 E(et)-.11 E(sizes is that the pack)100.8 675 Q
(et size itself can in\215uence the pack)-.11 E
(et drop rate [V00, WBL04].)-.11 E(Flo)72 733 Q(yd/K)-.11 E 268.584
(ohler Section)-.385 F 2.75(B.3. [P)2.75 F(age 25])-.165 E 0 Cg EP
%%Page: 26 26
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(The def)100.8 85 Q
(ault TFRC w)-.11 E(as designed for rough f)-.11 E(airness with TCP)-.11
E 2.75(,f)-1.221 G(or TFRC and TCP \215o)-2.75 E(ws)-.275 E
(with the same pack)100.8 98 Q(et size and e)-.11 E
(xperiencing the same pack)-.165 E(et drop rate.)-.11 E
(When the issue of)5.5 E -.11(fa)100.8 111 S(irness between \215o).11 E
(ws with dif)-.275 E(ferent pack)-.275 E
(ets sizes is addressed, it matters whether the)-.11 E(pack)100.8 124 Q
(et drop rates e)-.11 E(xperienced by the \215o)-.165 E
(ws is related to the pack)-.275 E(et size.)-.11 E(That is, are small)
5.5 E(pack)100.8 137 Q(ets just as lik)-.11 E(ely to be dropped as lar)
-.11 E(ge TCP pack)-.198 E(ets, or are the smaller pack)-.11 E(ets less)
-.11 E(lik)100.8 150 Q(ely to be dropped [WBL04].)-.11 E
(And what is the relationship between the pack)5.5 E(et-dropping)-.11 E
(beha)100.8 163 Q(vior of the path, and the loss e)-.22 E -.165(ve)-.275
G(nt measurements of TFRC?).165 E/F1 11/Courier@0 SF 6.6(<-----S)213 182
S(end Rates in Kbps - - - - >)-6.6 E 46.2(Web TCP)153.6 195 R
(\(1460B seg\))6.6 E(TFRC-SP \(200B seg\))33 E 13.2(Sessions DropRate)
140.4 208 R 19.8(SendRate DropRate)13.2 F(SendRate)13.2 E 13.2
(-------- --------)140.4 221 R 19.8(-------- --------)13.2 F(--------)
13.2 E 39.6(10 0.04)166.8 234 R 39.6(316.18 0.05)33 F(183.05)33 E 39.6
(25 0.07)166.8 247 R 39.6(227.47 0.07)33 F(181.23)33 E 39.6(50 0.08)
166.8 260 R 39.6(181.10 0.08)33 F(178.32)33 E 39.6(100 0.14)160.2 273 R
39.6(85.97 0.12)39.6 F(151.42)33 E 39.6(200 0.17)160.2 286 R 39.6
(61.20 0.14)39.6 F(73.88)39.6 E 39.6(400 0.20)160.2 299 R 39.6
(27.79 0.18)39.6 F(36.81)39.6 E 39.6(800 0.29 3.50 0.27)160.2 312 R
(16.33)39.6 E 39.6(1600 0.37 0.63 0.33 6.29)153.6 325 R
(Table 12: Drop and Send Rates for Drop-Tail Queues in Packets)127.2 351
Q F0 -.88(Ta)100.8 373.6 S(ble 12 sho).88 E(ws the results of the secon\
d half of 100-second simulations, with \214v)-.275 E 2.75(eT)-.165 G(CP)
-2.75 E(connections and \214v)100.8 386.6 Q 2.75(eT)-.165 G
(FRC-SP connections competing with web traf)-2.75 E
(\214c in a topology with)-.275 E 2.75(a3M)100.8 399.6 S
(bps shared link.)-2.75 E
(The TFRC-SP application generates 200-byte data pack)5.5 E(ets e)-.11 E
-.165(ve)-.275 G(ry 10).165 E(ms, for a maximum data rate of 160 Kbps.)
100.8 412.6 Q(The \214v)5.5 E 2.75(el)-.165 G(ong-li)-2.75 E -.165(ve)
-.275 G 2.75(dT).165 G(CP connections use a data)-2.75 E(pack)100.8
425.6 Q(et size of 1460 bytes, and the web traf)-.11 E
(\214c uses a data pack)-.275 E(et size of 512 bytes.)-.11 E(The \214v)
5.5 E(e)-.165 E(TCP connections ha)100.8 438.6 Q .33 -.165(ve r)-.22 H
(oundtrip times from 40 to 240 ms, and the \214v).165 E 2.75(eT)-.165 G
(FRC connections)-2.75 E(ha)100.8 451.6 Q .33 -.165(ve t)-.22 H
(he same set of round-trip times.).165 E(The SA)5.5 E
(CK TCP connections in these simulations)-.44 E(use the def)100.8 464.6
Q(ault parameters in the NS simulator)-.11 E 2.75(,w)-.44 G
(ith Limited T)-2.75 E(ransmit, and a minimum)-.385 E -1.628 -.66(RT O)
100.8 477.6 T(of 200 ms.)3.41 E
(Adding timestamps to the TCP connection didn')5.5 E 2.75(tc)-.198 G
(hange the results)-2.75 E(appreciably)100.8 490.6 Q 5.5(.T)-.715 G
(he simulations include re)-5.5 E -.165(ve)-.275 G(rse-path traf).165 E
(\214c, to add some small control)-.275 E(pack)100.8 503.6 Q
(ets to the forw)-.11 E(ard path, and some queueing delay to the re)-.11
E -.165(ve)-.275 G(rse path.).165 E(The number of)5.5 E
(web sessions is v)100.8 516.6 Q(aried to create dif)-.275 E(ferent le)
-.275 E -.165(ve)-.275 G(ls of congestion.).165 E(The Drop-T)5.5 E
(ail queue is in)-.88 E(units of pack)100.8 529.6 Q
(ets, which each pack)-.11 E(et arri)-.11 E
(ving to the queue requires a single b)-.275 E(uf)-.22 E(fer)-.275 E
2.75(,r)-.44 G -2.475 -.165(eg a)-2.75 H(rdless).165 E(of the pack)100.8
542.6 Q(et size.)-.11 E -.88(Ta)100.8 559.2 S(ble 12 sho).88 E(ws the a)
-.275 E -.165(ve)-.22 G(rage TCP and TFRC sending rates, each a).165 E
-.165(ve)-.22 G(raged o).165 E -.165(ve)-.165 G 2.75(rt).165 G(he \214v)
-2.75 E(e)-.165 E<8d6f>100.8 572.2 Q 2.75(ws. As)-.275 F -.165(ex)2.75 G
(pected, the TFRC-SP \215o).165 E(ws see similar pack)-.275 E
(et drop rates as the TCP \215o)-.11 E(ws,)-.275 E
(though the TFRC-SP \215o)100.8 585.2 Q(ws recei)-.275 E -.165(ve)-.275
G 2.75(sh).165 G(igher throughput than the TCP \215o)-2.75 E
(ws with pack)-.275 E(et drop)-.11 E(rates of 25% or higher)100.8 598.2
Q(.)-.605 E(Flo)72 733 Q(yd/K)-.11 E 268.584(ohler Section)-.385 F 2.75
(B.3. [P)2.75 F(age 26])-.165 E 0 Cg EP
%%Page: 27 27
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<-----S)226.2 85 S(end Rates in Kbps - - - - >)-6.6 E 39.6(Web TCP)
166.8 98 R(\(1460B seg\))6.6 E(TFRC-SP \(200B seg\))39.6 E 13.2
(Sessions DropRate)153.6 111 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)153.6 124 R 19.8(-------- --------)13.2 F
(--------)13.2 E 33(10 0.061)180 137 R 33(239.81 0.004)33 F(185.19)33 E
33(25 0.089)180 150 R 33(189.02 0.006)33 F(184.95)33 E 33
(50 0.141 99.46 0.013)180 163 R(185.07)33 E 33(100 0.196 16.42 0.022)
173.4 176 R(183.77)33 E 33(200 0.256)173.4 189 R 33(4.46 0.032)46.2 F
(181.98)33 E 33(400 0.291)173.4 202 R 33(4.61 0.051)46.2 F(151.88)33 E
33(800 0.487)173.4 215 R 33(1.01 0.078)46.2 F(113.10)33 E 33(1600 0.648)
166.8 228 R 33(0.67 0.121 65.17)46.2 F
(Table 13: Drop and Send Rates for Drop-Tail Queues in Bytes I:)107.4
267 Q(1460B TCP Segments)272.4 280 Q F0(Ho)100.8 302.6 Q(we)-.275 E
-.165(ve)-.275 G .88 -.44(r, t).165 H(he f).44 E
(airness results can change signi\214cantly if the Drop-T)-.11 E
(ail queue at the)-.88 E
(congested output link is in units of bytes rather than pack)100.8 315.6
Q 2.75(ets. F)-.11 F(or a queue in pack)-.165 E(ets, the)-.11 E
(queue has a \214x)100.8 328.6 Q(ed number of b)-.165 E(uf)-.22 E
(fers, and each b)-.275 E(uf)-.22 E(fer can hold e)-.275 E
(xactly one pack)-.165 E(et,)-.11 E(re)100.8 341.6 Q -.055(ga)-.165 G
(rdless of its size in bytes.).055 E -.165(Fo)5.5 G 2.75(raq).165 G
(ueue in bytes, the queue has a \214x)-2.75 E(ed number of)-.165 E
(*bytes*, and an almost-full queue might ha)100.8 354.6 Q .33 -.165
(ve r)-.22 H(oom for a small pack).165 E(et b)-.11 E(ut not for a lar)
-.22 E(ge)-.198 E 2.75(one. This,)100.8 367.6 R
(for a simulation with a Drop-T)2.75 E(ail queue in bytes, lar)-.88 E
(ge pack)-.198 E(ets are more lik)-.11 E(ely)-.11 E
(to be dropped than are small ones.)100.8 380.6 Q
(The NS simulator doesn')5.5 E 2.75(ty)-.198 G(et ha)-2.75 E .33 -.165
(ve a m)-.22 H(ore realistic).165 E
(intermediate model, where the queue has a \214x)100.8 393.6 Q
(ed number of b)-.165 E(uf)-.22 E(fers, each b)-.275 E(uf)-.22 E
(fer has a \214x)-.275 E(ed)-.165 E(number of bytes, and each pack)100.8
406.6 Q(et w)-.11 E(ould require one or more free b)-.11 E(uf)-.22 E
2.75(fers. In)-.275 F(this model, a)2.75 E(small pack)100.8 419.6 Q
(et w)-.11 E(ould use one b)-.11 E(uf)-.22 E(fer)-.275 E 2.75(,w)-.44 G
(hile a lar)-2.75 E(ger pack)-.198 E(et w)-.11 E(ould require se)-.11 E
-.165(ve)-.275 G(ral b).165 E(uf)-.22 E(fers.)-.275 E
(The scenarios in T)100.8 436.2 Q(able 13 are identical to those in T)
-.88 E(able 12, e)-.88 E(xcept that the Drop-T)-.165 E(ail queue)-.88 E
(is in units of bytes instead of pack)100.8 449.2 Q 2.75(ets. Thus,)-.11
F<8c76>2.75 E 2.75(eT)-.165 G(CP connections and \214v)-2.75 E 2.75(eT)
-.165 G(FRC-SP)-2.75 E(connections compete with web traf)100.8 462.2 Q
(\214c in a topology with a 3 Mbps shared link, with each)-.275 E
(TFRC-SP application generating 200-byte data pack)100.8 475.2 Q(ets e)
-.11 E -.165(ve)-.275 G(ry 10 ms, for a maximum data).165 E
(rate of 160 Kbps.)100.8 488.2 Q(The number of web sessions is v)5.5 E
(aried to create dif)-.275 E(ferent le)-.275 E -.165(ve)-.275 G(ls of)
.165 E 2.75(congestion. Ho)100.8 501.2 R(we)-.275 E -.165(ve)-.275 G .88
-.44(r, i).165 H(nstead of Drop-T).44 E
(ail queues able to accommodate at most a hundred)-.88 E(pack)100.8
514.2 Q(ets, the routers' Drop-T)-.11 E
(ail queues are each able to accommodate at most 50,000 bytes)-.88 E
(\(computed in NS as a hundred pack)100.8 527.2 Q
(ets times the mean pack)-.11 E(etsize of 500 bytes\).)-.11 E(As T)100.8
543.8 Q(able 13 sho)-.88 E(ws, with a Drop-T)-.275 E
(ail queue in bytes, the TFRC-SP \215o)-.88 E 2.75(ws)-.275 G
(ees a much)-2.75 E(smaller pack)100.8 556.8 Q
(et drop rate than the TCP \215o)-.11 E 1.43 -.715(w, a)-.275 H
(nd as a consequence recei).715 E -.165(ve)-.275 G 2.75(sam).165 G
(uch lar)-2.75 E(ger)-.198 E(sending rate.)100.8 569.8 Q -.165(Fo)5.5 G
2.75(rt).165 G(he simulations in T)-2.75 E(able 13, the TFRC-SP \215o)
-.88 E(ws use 200-byte data)-.275 E(se)100.8 582.8 Q
(gments, while the long-li)-.165 E -.165(ve)-.275 G 2.75(dT).165 G
(CP \215o)-2.75 E(ws use 1460-byte data se)-.275 E 2.75(gments. F)-.165
F(or e)-.165 E(xample,)-.165 E(when the \214v)100.8 595.8 Q 2.75(eT)
-.165 G(CP \215o)-2.75 E(ws and \214v)-.275 E 2.75(eT)-.165 G
(FRC-SP \215o)-2.75 E(ws share the link with 800 web sessions, the)-.275
E<8c76>100.8 608.8 Q 2.75(eT)-.165 G(CP \215o)-2.75 E(ws see an a)-.275
E -.165(ve)-.22 G
(rage drop rate of 49% in the second half of the simulation, while).165
E(the \214v)100.8 621.8 Q 2.75(eT)-.165 G(FRC-SP \215o)-2.75 E(ws recei)
-.275 E .33 -.165(ve a)-.275 H 2.75(na).165 G -.165(ve)-2.97 G
(rage drop rate of 8%, and as a consequence recei).165 E -.165(ve)-.275
G(more than 100 times the throughput of the TCP \215o)100.8 634.8 Q 2.75
(ws. This)-.275 F(raises serious questions about)2.75 E
(making the assumption that \215o)100.8 647.8 Q(ws with small pack)-.275
E(ets see the same pack)-.11 E(et drop rate as \215o)-.11 E(ws)-.275 E
(with lar)100.8 660.8 Q(ger pack)-.198 E 2.75(ets. Further)-.11 F -.11
(wo)2.75 G(rk will ha).11 E .33 -.165(ve t)-.22 H 2.75(oi).165 G
(nclude an in)-2.75 E -.165(ve)-.44 G(stig).165 E
(ation into the range of)-.055 E
(realistic Internet scenarios, in terms of whether lar)100.8 673.8 Q
(ge pack)-.198 E(ets are considerably more lik)-.11 E(ely)-.11 E
(to be dropped than are small ones.)100.8 686.8 Q(Flo)72 733 Q(yd/K)-.11
E 268.584(ohler Section)-.385 F 2.75(B.3. [P)2.75 F(age 27])-.165 E 0 Cg
EP
%%Page: 28 28
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<-----S)226.2 85 S(end Rates in Kbps - - - - >)-6.6 E 46.2(Web TCP)
166.8 98 R(\(512B seg\))6.6 E(TFRC-SP \(200B seg\))39.6 E 13.2
(Sessions DropRate)153.6 111 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)153.6 124 R 19.8(-------- --------)13.2 F
(--------)13.2 E 39.6(10 0.02)180 137 R 39.6(335.05 0.00)33 F(185.16)33
E 39.6(25 0.02)180 150 R 39.6(289.94 0.00)33 F(185.36)33 E 39.6(50 0.04)
180 163 R 39.6(139.99 0.01)33 F(184.98)33 E 39.6(100 0.06)173.4 176 R
39.6(53.50 0.01)39.6 F(184.66)33 E 39.6(200 0.10)173.4 189 R 39.6
(16.14 0.04)39.6 F(167.87)33 E 39.6(400 0.16 6.36 0.07)173.4 202 R
(114.85)33 E 39.6(800 0.24 0.90 0.11)173.4 215 R(67.23)39.6 E 39.6
(1600 0.42 0.35 0.18)166.8 228 R(39.32)39.6 E
(Table 14: Drop and Send Rates for Drop-Tail Queues in Bytes II:)107.4
254 Q(512B TCP Segments)279 267 Q F0 -.88(Ta)100.8 289.6 S(ble 14 sho)
.88 E(ws that in the scenario the long-li)-.275 E -.165(ve)-.275 G 2.75
(dT).165 G(CP \215o)-2.75 E(ws recei)-.275 E .33 -.165(ve a h)-.275 H
(igher pack).165 E(et drop)-.11 E(rate than the TFRC-SP \215o)100.8
302.6 Q(ws, and recei)-.275 E .33 -.165(ve c)-.275 H
(onsiderably less throughput, e).165 E -.165(ve)-.275 G 2.75(nw).165 G
(hen the)-2.75 E(long-li)100.8 315.6 Q -.165(ve)-.275 G 2.75(dT).165 G
(CP \215o)-2.75 E(ws use 512-byte se)-.275 E(gments.)-.165 E 1.76 -.88
(To s)100.8 332.2 T(ho).88 E 2.75(wt)-.275 G(he potential ne)-2.75 E
-.055(ga)-.165 G(ti).055 E .33 -.165(ve e)-.275 H -.275(ff).165 G
(ect of TFRC-SP in such an en).275 E(vironment, we consider a)-.44 E
(simulation with N TCP \215o)100.8 345.2 Q(ws, N TFRC-SP \215o)-.275 E
(ws, and 10*N web sessions, for N ranging)-.275 E
(from 1 to 50, so that the demand increases from all types of traf)100.8
358.2 Q(\214c, with routers with Drop-)-.275 E -.88(Ta)100.8 371.2 S
(il queues in bytes.).88 E F1 6.6(<-----S)226.2 390.2 S
(end Rates in Kbps - - - - >)-6.6 E 39.6(Web TCP)166.8 403.2 R
(\(1460B seg\))6.6 E(TFRC-SP \(200B seg\))39.6 E 13.2(Sessions DropRate)
153.6 416.2 R 19.8(SendRate DropRate)13.2 F(SendRate)13.2 E 13.2
(-------- --------)153.6 429.2 R 19.8(-------- --------)13.2 F(--------)
13.2 E 33(10 0.014)180 442.2 R 33(2085.36 0.001)26.4 F(180.29)33 E 33
(20 0.040)180 455.2 R 33(788.88 0.004)33 F(183.87)33 E 33(30 0.074)180
468.2 R 33(248.80 0.006)33 F(182.94)33 E 33(40 0.113)180 481.2 R 33
(163.20 0.008)33 F(185.33)33 E 33(50 0.127 94.70 0.011)180 494.2 R
(185.14)33 E 33(60 0.177 53.24 0.015)180 507.2 R(185.30)33 E 33
(70 0.174 35.31 0.016)180 520.2 R(185.07)33 E 33(80 0.221 19.38 0.019)
180 533.2 R(183.91)33 E 33(90 0.188 15.63 0.019)180 546.2 R(180.42)33 E
33(100 0.265)173.4 559.2 R 33(7.08 0.023)46.2 F(176.71)33 E 33
(200 0.324)173.4 572.2 R 33(0.38 0.042)46.2 F(139.48)33 E 33(300 0.397)
173.4 585.2 R 33(0.32 0.076 93.53)46.2 F 33(400 0.529)173.4 598.2 R 33
(0.40 0.100 70.40)46.2 F 33(500 0.610)173.4 611.2 R 33(0.37 0.121 57.59)
46.2 F(Table 15: Drop and Send Rates for Drop-Tail Queues in Bytes III:)
107.4 637.2 Q(TFRC-SP, 1460B TCP Segments)219.6 650.2 Q F0(Flo)72 733 Q
(yd/K)-.11 E 268.584(ohler Section)-.385 F 2.75(B.3. [P)2.75 F(age 28])
-.165 E 0 Cg EP
%%Page: 29 29
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<-----S)226.2 85 S(end Rates in Kbps - - - - >)-6.6 E 39.6(Web TCP)
166.8 98 R(\(1460B seg\))6.6 E(TFRC \(200B seg\))46.2 E 13.2
(Sessions DropRate)153.6 111 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)153.6 124 R 19.8(-------- --------)13.2 F
(--------)13.2 E 33(10 0.016)180 137 R 33(1926.00 0.002)26.4 F(178.47)33
E 33(20 0.030)180 150 R 33(805.20 0.003)33 F(178.23)33 E 33(30 0.062)180
163 R 33(346.24 0.005)33 F(161.19)33 E 33(40 0.093)180 176 R 33
(219.18 0.007)33 F(136.28)33 E 33(50 0.110)180 189 R 33
(132.77 0.010 83.02)33 F 33(60 0.170 88.88 0.014 69.75)180 202 R 33
(70 0.149 70.73 0.015 55.59)180 215 R 33(80 0.213 43.17 0.020 42.82)180
228 R 33(90 0.188 36.19 0.017 43.61)180 241 R 33
(100 0.233 24.77 0.026 35.17)173.4 254 R 33(200 0.311)173.4 267 R 33
(5.46 0.034 24.85)46.2 F 33(300 0.367)173.4 280 R 33(3.74 0.049 20.19)
46.2 F 33(400 0.421)173.4 293 R 33(2.59 0.055 17.71)46.2 F 33(500 0.459)
173.4 306 R 33(1.10 0.069 15.95)46.2 F
(Table 16: Drop and Send Rates for Drop-Tail Queues in Bytes IV:)107.4
319 Q(Standard TFRC, 1460B TCP Segments)199.8 332 Q F0 -.88(Ta)100.8
354.6 S(ble 15 sho).88 E(ws simulations using TFRC-SP)-.275 E 2.75(,w)
-1.221 G(hile T)-2.75 E(able 16 sho)-.88 E(ws simulations using)-.275 E
(TFRC instead of TFRC-SP)100.8 367.6 Q 5.5(.T)-1.221 G
(his is the only dif)-5.5 E(ference between the simulations in the tw)
-.275 E(o)-.11 E 2.75(tables. Note)100.8 380.6 R
(that when TFRC-SP is used, the TCP \215o)2.75 E(ws and web traf)-.275 E
(\214c are essentially)-.275 E(starv)100.8 393.6 Q
(ed, while the TFRC-SP \215o)-.165 E(ws each continue to send 57 Kbps.)
-.275 E(In contrast, when)5.5 E
(standard TFRC is used instead of TFRC-SP for the \215o)100.8 406.6 Q
(ws sending 200-byte se)-.275 E(gments, the)-.165 E(TCP \215o)100.8
419.6 Q(ws are not starv)-.275 E(ed \(although the)-.165 E 2.75(ys)-.165
G(till don')-2.75 E 2.75(tr)-.198 G(ecei)-2.75 E .33 -.165(ve t)-.275 H
(heir "share" of the link).165 E(bandwidth when their pack)100.8 432.6 Q
(et drop rates are 30% or higher)-.11 E 2.75(.\) These)-.605 F(tw)2.75 E
2.75(os)-.11 G(ets of simulations)-2.75 E
(illustrate the dangers of a widespread deplo)100.8 445.6 Q
(yment of TFRC-SP in an en)-.11 E(vironment where)-.44 E(small pack)
100.8 458.6 Q(ets are less lik)-.11 E(ely to be dropped than lar)-.11 E
(ge ones.)-.198 E/F2 11/Times-Bold@0 SF(B.4.)72 484.6 Q/F3 13
/Times-Bold@0 SF -.13(Pa)5.5 G(ck).13 E(et Dr)-.13 E(opping Beha)-.234 E
(vior at Routers with A)-.325 E(QM)-.585 E F0(As e)100.8 501.2 Q
(xpected, the pack)-.165 E(et dropping beha)-.11 E(vior also can be v)
-.22 E(aried by v)-.275 E(arying the Acti)-.275 E .33 -.165(ve Q)-.275 H
(ueue).165 E(Management \(A)100.8 514.2 Q(QM\) mechanism in the router)
-.605 E 5.5(.W)-.605 G(hen the routers use RED \(Random Early)-5.5 E
(Detection\), there are se)100.8 527.2 Q -.165(ve)-.275 G
(ral parameters than can af).165 E(fect the pack)-.275 E
(et dropping or marking)-.11 E(beha)100.8 540.2 Q
(vior as a function of the pack)-.22 E(et size.)-.11 E
(First, as with Drop-T)100.8 556.8 Q
(ail, the RED queue can be either in units of pack)-.88 E
(ets or of bytes.)-.11 E(This)5.5 E(can af)100.8 569.8 Q(fect the pack)
-.275 E(et dropping beha)-.11 E
(vior when RED is unable to control the a)-.22 E -.165(ve)-.22 G
(rage queue).165 E(size, and the queue o)100.8 582.8 Q -.165(ve)-.165 G
(r\215o).165 E(ws.)-.275 E(Second, and orthogonally)100.8 599.4 Q 2.75
(,R)-.715 G(ED can be con\214gured to be either in pack)-2.75 E
(et mode or in byte)-.11 E 2.75(mode. In)100.8 612.4 R(pack)2.75 E
(et mode, each *pack)-.11 E
(et* has the same probability of being dropped by RED,)-.11 E(while in \
byte mode, each *byte* has the same probability of being dropped.)100.8
625.4 Q(In pack)5.5 E(et)-.11 E(mode, lar)100.8 638.4 Q(ge-pack)-.198 E
(et and small-pack)-.11 E(et \215o)-.11 E(ws recei)-.275 E .33 -.165
(ve r)-.275 H(oughly the same pack).165 E(et drop rate,)-.11 E
(while in byte mode, lar)100.8 651.4 Q(ge-pack)-.198 E
(et and small-pack)-.11 E(et \215o)-.11 E
(ws with the same throughput in bps)-.275 E(recei)100.8 664.4 Q .33
-.165(ve r)-.275 H(oughly the same *number* of pack).165 E(et drops.)
-.11 E([EA03] assessed the impact of byte)5.5 E(vs. pack)100.8 677.4 Q
(et modes on RED performance.)-.11 E(Flo)72 733 Q(yd/K)-.11 E 268.584
(ohler Section)-.385 F 2.75(B.4. [P)2.75 F(age 29])-.165 E 0 Cg EP
%%Page: 30 30
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(The simulations reported belo)
100.8 85 Q 2.75(ws)-.275 G(ho)-2.75 E 2.75(wt)-.275 G
(hat for RED in pack)-2.75 E(et mode, the pack)-.11 E(et drop rates)-.11
E(for the TFRC-SP \215o)100.8 98 Q
(ws are similar to those for the TCP \215o)-.275 E
(ws, with a resulting acceptable)-.275 E
(throughput for the TFRC-SP \215o)100.8 111 Q 5.5(ws. This)-.275 F
(is true with the queue in pack)2.75 E(ets or in bytes, and)-.11 E
(with or without Adapti)100.8 124 Q .33 -.165(ve R)-.275 H
(ED \(discussed belo).165 E 2.75(w\). As)-.275 F(we sho)2.75 E 2.75(wb)
-.275 G(elo)-2.75 E 1.43 -.715(w, t)-.275 H(his f).715 E(airness)-.11 E
(between TCP and TFRC-SP \215o)100.8 137 Q
(ws does not hold for RED in byte mode.)-.275 E
(The third RED parameter that af)100.8 153.6 Q(fects the pack)-.275 E
(et dropping or marking beha)-.11 E(vior as a)-.22 E(function of pack)
100.8 166.6 Q
(et size is whether the RED mechanism is using Standard RED or Adapti)
-.11 E -.165(ve)-.275 G 2.75(RED; Adapti)100.8 179.6 R .33 -.165(ve R)
-.275 H(ED tries to maintain the same a).165 E -.165(ve)-.22 G
(rage queue size, re).165 E -.055(ga)-.165 G(rdless of the).055 E(pack)
100.8 192.6 Q(et drop rate.)-.11 E(The use of Adapti)5.5 E .33 -.165
(ve R)-.275 H(ED allo).165 E(ws the RED mechanism to function more)-.275
E(ef)100.8 205.6 Q(fecti)-.275 E -.165(ve)-.275 G
(ly in the presence of high pack).165 E
(et drop rates \(e.g., greater than 10%\).)-.11 E -.44(Wi)5.5 G(thout)
.44 E(Adapti)100.8 218.6 Q .33 -.165(ve R)-.275 H(ED, there is a \214x)
.165 E(ed dropping threshold, and all arri)-.165 E(ving pack)-.275 E
(ets are dropped)-.11 E(when the dropping or marking rate e)100.8 231.6
Q(xceeds this threshold.)-.165 E(In contrast, with Adapti)5.5 E .33
-.165(ve R)-.275 H(ED,).165 E
(the dropping function is adapted to accommodate high-drop-rate re)100.8
244.6 Q 2.75(gimes. One)-.165 F
(consequence is that when byte mode is used with Adapti)100.8 257.6 Q
.33 -.165(ve R)-.275 H(ED, the byte mode e).165 E(xtends)-.165 E -2.365
-.275(ev e)100.8 270.6 T 2.75(nt).275 G 2.75(oh)-2.75 G
(igh-drop-rate re)-2.75 E 2.75(gimes. When)-.165 F
(byte mode is used with standard RED, ho)2.75 E(we)-.275 E -.165(ve)
-.275 G .88 -.44(r, t).165 H(he).44 E
(byte mode is no longer in use when the drop rate e)100.8 283.6 Q
(xceeds the \214x)-.165 E(ed dropping threshold \(set)-.165 E(by def)
100.8 296.6 Q(ault to 10% in the NS simulator\).)-.11 E
(In the simulations in this section, we e)100.8 313.2 Q
(xplore the TFRC-SP beha)-.165 E(vior o)-.22 E -.165(ve)-.165 G 2.75(rs)
.165 G(ome of this)-2.75 E(range of scenarios.)100.8 326.2 Q
(In this simulations, as in Section B.3 abo)5.5 E -.165(ve)-.165 G 2.75
(,t).165 G(he application for the)-2.75 E(TFRC-SP \215o)100.8 339.2 Q
2.75(wu)-.275 G(ses 200-byte data pack)-2.75 E(ets, generating 100 pack)
-.11 E(ets per second.)-.11 E/F1 11/Courier@0 SF 6.6(<-----S)226.2 358.2
S(end Rates in Kbps - - - - >)-6.6 E 46.2(Web TCP)166.8 371.2 R
(\(1460B seg\))6.6 E(TFRC-SP \(200B seg\))33 E 13.2(Sessions DropRate)
153.6 384.2 R 19.8(SendRate DropRate)13.2 F(SendRate)13.2 E 13.2
(-------- --------)153.6 397.2 R 19.8(-------- --------)13.2 F(--------)
13.2 E 39.6(10 0.05)180 410.2 R 39.6(305.76 0.04)33 F(182.82)33 E 39.6
(25 0.06)180 423.2 R 39.6(224.16 0.06)33 F(175.91)33 E 39.6(50 0.09)180
436.2 R 39.6(159.12 0.08)33 F(152.51)33 E 39.6(100 0.13)173.4 449.2 R
39.6(90.77 0.11)39.6 F(106.13)33 E 39.6(200 0.14)173.4 462.2 R 39.6
(48.53 0.14)39.6 F(70.25)39.6 E 39.6(400 0.20)173.4 475.2 R 39.6
(22.08 0.19)39.6 F(41.50)39.6 E 39.6(800 0.27 3.55 0.25)173.4 488.2 R
(17.50)39.6 E 39.6(1600 0.42 1.87 0.34 8.81)166.8 501.2 R
(Table 17: Drop and Send Rates for RED Queues in Packet Mode)114 527.2 Q
F0 -.165(Fo)100.8 549.8 S 2.75(rt).165 G(he simulations in T)-2.75 E
(able 17, with a congested router with a RED queue in pack)-.88 E
(et mode,)-.11 E(the results are similar to those with a Drop-T)100.8
562.8 Q(ail queue in pack)-.88 E(ets, as in T)-.11 E(able 12 abo)-.88 E
-.165(ve)-.165 G(.).165 E(The TFRC-SP \215o)100.8 575.8 Q 2.75(wr)-.275
G(ecei)-2.75 E -.165(ve)-.275 G 2.75(ss).165 G(imilar pack)-2.75 E
(et drop rates as the TCP \215o)-.11 E 1.43 -.715(w, t)-.275 H
(hough it recei).715 E -.165(ve)-.275 G(s).165 E
(higher throughput in the more congested en)100.8 588.8 Q 2.75
(vironments. The)-.44 F(simulations are similar with a)2.75 E
(RED queue in pack)100.8 601.8 Q
(et mode with the queue in bytes, and with or without Adapti)-.11 E .33
-.165(ve R)-.275 H 2.75(ED. In).165 F(these simulations, TFRC-SP gi)
100.8 614.8 Q -.165(ve)-.275 G 2.75(sr).165 G
(oughly the desired performance.)-2.75 E(Flo)72 733 Q(yd/K)-.11 E
268.584(ohler Section)-.385 F 2.75(B.4. [P)2.75 F(age 30])-.165 E 0 Cg
EP
%%Page: 31 31
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<-----S)226.2 85 S(end Rates in Kbps - - - - >)-6.6 E 39.6(Web TCP)
166.8 98 R(\(1460B seg\))6.6 E(TFRC-SP \(200B seg\))39.6 E 13.2
(Sessions DropRate)153.6 111 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)153.6 124 R 19.8(-------- --------)13.2 F
(--------)13.2 E 39.6(10 0.06)180 137 R 39.6(272.16 0.02)33 F(184.37)33
E 39.6(25 0.07)180 150 R 39.6(175.82 0.02)33 F(184.06)33 E 39.6(50 0.10)
180 163 R 39.6(75.65 0.04)39.6 F(180.56)33 E 39.6(100 0.14)173.4 176 R
39.6(38.98 0.07)39.6 F(151.65)33 E 39.6(200 0.19)173.4 189 R 39.6
(16.66 0.11)39.6 F(106.80)33 E 39.6(400 0.26 4.85 0.15)173.4 202 R
(69.41)39.6 E 39.6(800 0.35 3.12 0.20)173.4 215 R(27.07)39.6 E 39.6
(1600 0.42 0.67 0.29)166.8 228 R(10.68)39.6 E
(Table 18: Drop and Send Rates for RED Queues in Byte Mode)127.2 254 Q
F0 -.88(Ta)100.8 276.6 S(ble 18 sho).88 E
(ws that with a standard RED queue in byte mode instead of pack)-.275 E
(et mode, there)-.11 E(is a some)100.8 289.6 Q(what greater dif)-.275 E
(ferent between the pack)-.275 E
(et drop rates between the TCP and TFRC-)-.11 E(SP \215o)100.8 302.6 Q
(ws, particularly for lo)-.275 E(wer pack)-.275 E(et drop rates.)-.11 E
-.165(Fo)5.5 G 2.75(rt).165 G(he simulation in T)-2.75 E(able 18, the)
-.88 E(pack)100.8 315.6 Q(et drop rates for the TCP \215o)-.11 E
(ws can range from 1 1/2 to four times greater than the)-.275 E(pack)
100.8 328.6 Q(et drop rates for the TFRC-SP \215o)-.11 E 2.75(ws. Ho)
-.275 F(we)-.275 E -.165(ve)-.275 G .88 -.44(r, b).165 H
(ecause the TFRC-SP \215o).44 E 2.75(wh)-.275 G(as an)-2.75 E
(upper bound on the sending rate, its sending rate is not af)100.8 341.6
Q(fected in the lo)-.275 E(wer pack)-.275 E(et-drop-)-.11 E(rate re)
100.8 354.6 Q(gimes; its sending rate is only af)-.165 E
(fected in the re)-.275 E(gimes with pack)-.165 E(et drop rates of 10%)
-.11 E(or more.)100.8 367.6 Q
(The sending rate for TFRC-SP in the scenarios in T)5.5 E
(able 18 with higher pack)-.88 E(et)-.11 E(drop rates are greater than \
desired, e.g., for the scenarios with 400 web sessions or greater)100.8
380.6 Q(.)-.605 E(Ho)100.8 393.6 Q(we)-.275 E -.165(ve)-.275 G .88 -.44
(r, t).165 H
(he results with TFRC-SP are at least better than that of small-pack).44
E(et \215o)-.11 E(ws with)-.275 E(no congestion control at all.)100.8
406.6 Q F1 6.6(<-----S)226.2 425.6 S(end Rates in Kbps - - - - >)-6.6 E
46.2(Web TCP)166.8 438.6 R(\(512B seg\))6.6 E(TFRC-SP \(200B seg\))39.6
E 13.2(Sessions DropRate)153.6 451.6 R 19.8(SendRate DropRate)13.2 F
(SendRate)13.2 E 13.2(-------- --------)153.6 464.6 R 19.8
(-------- --------)13.2 F(--------)13.2 E 39.6(10 0.01)180 477.6 R 39.6
(337.86 0.01)33 F(184.06)33 E 39.6(25 0.02)180 490.6 R 39.6(258.71 0.01)
33 F(184.03)33 E 39.6(50 0.02)180 503.6 R 39.6(184.71 0.01)33 F(183.99)
33 E 39.6(100 0.04)173.4 516.6 R 39.6(63.63 0.03)39.6 F(184.43)33 E 39.6
(200 0.08)173.4 529.6 R 39.6(28.95 0.06)39.6 F(149.80)33 E 39.6
(400 0.12)173.4 542.6 R 39.6(17.03 0.10)39.6 F(88.21)39.6 E 39.6
(800 0.24 5.94 0.15)173.4 555.6 R(36.80)39.6 E 39.6(1600 0.32 3.37 0.21)
166.8 568.6 R(19.45)39.6 E
(Table 19: Drop and Send Rates for RED Queues in Byte Mode)127.2 594.6 Q
F0 -.88(Ta)100.8 617.2 S(ble 19 sho).88 E
(ws that with a standard RED queue in byte mode and with long-li)-.275 E
-.165(ve)-.275 G 2.75(dT).165 G(CP)-2.75 E<8d6f>100.8 630.2 Q
(ws with 512-byte data se)-.275 E(gments, there is only a moderate dif)
-.165 E(ference between the pack)-.275 E(et)-.11 E
(drop rate for the 552-byte TCP pack)100.8 643.2 Q
(ets and the 240-byte TFRC-SP pack)-.11 E 2.75(ets. Ho)-.11 F(we)-.275 E
-.165(ve)-.275 G .88 -.44(r, t).165 H(he).44 E
(sending rate for TFRC-SP in the scenarios in T)100.8 656.2 Q
(able 19 with higher pack)-.88 E(et drop rates are)-.11 E
(still greater than desired, e)100.8 669.2 Q -.165(ve)-.275 G 2.75(ng)
.165 G -2.365 -.275(iv e)-2.75 H 2.75(nt).275 G(he goal of f)-2.75 E
(airness with TCP \215o)-.11 E(ws with 1500-byte)-.275 E(data se)100.8
682.2 Q(gments instead of 512-byte data se)-.165 E(gments.)-.165 E(Flo)
72 733 Q(yd/K)-.11 E 268.584(ohler Section)-.385 F 2.75(B.4. [P)2.75 F
(age 31])-.165 E 0 Cg EP
%%Page: 32 32
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<-----S)226.2 85 S(end Rates in Kbps - - - - >)-6.6 E 39.6(Web TCP)
166.8 98 R(\(1460B seg\))6.6 E(TFRC-SP \(200B seg\))39.6 E 13.2
(Sessions DropRate)153.6 111 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)153.6 124 R 19.8(-------- --------)13.2 F
(--------)13.2 E 39.6(10 0.04)180 137 R 39.6(318.10 0.02)33 F(185.34)33
E 39.6(25 0.08)180 150 R 39.6(175.34 0.03)33 F(184.38)33 E 39.6(50 0.10)
180 163 R 39.6(81.60 0.04)39.6 F(181.95)33 E 39.6(100 0.12)173.4 176 R
39.6(28.51 0.05)39.6 F(178.79)33 E 39.6(200 0.20 3.65 0.06)173.4 189 R
(173.78)33 E 39.6(400 0.27 1.44 0.08)173.4 202 R(161.41)33 E 39.6
(800 0.40 0.58 0.06)173.4 215 R(159.62)33 E 39.6(1600 0.55 0.29 0.02)
166.8 228 R(180.92)33 E
(Table 20: Drop and Send Rates with Adaptive RED Queues in Byte Mode)
100.8 254 Q F0 -.165(Fo)100.8 276.6 S 2.75(rt).165 G
(he simulations in T)-2.75 E
(able 20, the congested router uses an Adapti)-.88 E .33 -.165(ve R)
-.275 H(ED queue in byte).165 E(mode.)100.8 289.6 Q -.165(Fo)100.8 306.2
S 2.75(rt).165 G(his router)-2.75 E 2.75(,t)-.44 G
(he output queue is in units of bytes rather than of pack)-2.75 E
(ets, each *byte* is)-.11 E(dropped with the same probability)100.8
319.2 Q 2.75(,a)-.715 G(nd because of the use of Adapti)-2.75 E .33
-.165(ve R)-.275 H(ED, this byte-).165 E(dropping mode e)100.8 332.2 Q
(xtends e)-.165 E -.165(ve)-.275 G 2.75(nf).165 G(or the high-pack)-2.75
E(et-drop-rate re)-.11 E(gime.)-.165 E(As T)100.8 348.8 Q(able 20 sho)
-.88 E(ws, for a scenario with Adapti)-.275 E .33 -.165(ve R)-.275 H
(ED in byte mode, the pack).165 E(et drop rate)-.11 E
(for the TFRC-SP traf)100.8 361.8 Q(\214c is *much* lo)-.275 E
(wer than that for the TCP traf)-.275 E(\214c, and as a)-.275 E
(consequence, the sending rate for the TFRC-SP traf)100.8 374.8 Q
(\214c in a highly congested en)-.275 E(vironment is)-.44 E
(*much* higher than that of the TCP traf)100.8 387.8 Q 2.75(\214c. In)
-.275 F -.11(fa)2.75 G(ct, in these scenarios the TFRC-SP).11 E
(congestion control mechanisms are lar)100.8 400.8 Q(gely inef)-.198 E
(fecti)-.275 E .33 -.165(ve f)-.275 H(or the small-pack).165 E(et traf)
-.11 E(\214c.)-.275 E(The simulation with 1600 web serv)100.8 417.4 Q
(ers is of particular concern, because the TCP pack)-.165 E(et)-.11 E
(drop rate increases, while the TFRC-SP pack)100.8 430.4 Q
(et drop rate decreases signi\214cantly)-.11 E 5.5(.T)-.715 G(his is)
-5.5 E(due to a detail of the Adapti)100.8 443.4 Q .33 -.165(ve R)-.275
H(ED implementation in the NS simulator).165 E 2.75(,a)-.44 G
(nd not about the)-2.75 E(dynamics of TFRC-SP)100.8 456.4 Q 5.5(.I)
-1.221 G 2.75(np)-5.5 G(articular)-2.75 E 2.75(,A)-.44 G(dapti)-2.75 E
.33 -.165(ve R)-.275 H(ED is con\214gured not to "adapt" to pack).165 E
(et)-.11 E(drop rates o)100.8 469.4 Q -.165(ve)-.165 G 2.75(r5).165 G
2.75(0%. When)-2.75 F(the pack)2.75 E
(et drop rate is at most 50%, Adapti)-.11 E .33 -.165(ve R)-.275 H
(ED is).165 E(moderately successful in k)100.8 482.4 Q(eeping the pack)
-.11 E(et drop rate proportional to the pack)-.11 E(et size - TCP)-.11 E
(pack)100.8 495.4 Q(ets are six times lar)-.11 E
(ger than the TFRC-SP pack)-.198 E
(ets \(including headers\), so the TCP)-.11 E(pack)100.8 508.4 Q
(ets should see a pack)-.11 E(et drop rate roughly six times lar)-.11 E
(ger)-.198 E 5.5(.B)-.605 G(ut for pack)-5.5 E(et drop rates)-.11 E
-.165(ove)100.8 521.4 S 2.75(r5).165 G(0%, Adapti)-2.75 E .33 -.165
(ve R)-.275 H(ED is no longer in this re).165 E
(gime, so it is no longer successful in)-.165 E -.11(ke)100.8 534.4 S
(eping the drop rate for TCP pack).11 E
(ets at most six times the drop rate of the TFRC-SP)-.11 E(pack)100.8
547.4 Q(ets.)-.11 E 1.76 -.88(We n)100.8 564 T(ote that the unf).88 E
(airness in these simulations, in f)-.11 E -.22(avo)-.11 G 2.75(ro).22 G
2.75(fT)-2.75 G(FRC-SP)-2.75 E 2.75(,i)-1.221 G 2.75(se)-2.75 G -.165
(ve)-3.025 G 2.75(nm).165 G(ore se)-2.75 E -.165(ve)-.275 G(re).165 E
(than the unf)100.8 577 Q(airness sho)-.11 E(wn in T)-.275 E
(able 13 for a Drop-T)-.88 E(ail queue in bytes.)-.88 E
(At the same time, it)5.5 E(is not kno)100.8 590 Q(wn if there is an)
-.275 E 2.75(yd)-.165 G(eplo)-2.75 E(yment in the Internet of an)-.11 E
2.75(yr)-.165 G(outers with Adapti)-2.75 E .33 -.165(ve R)-.275 H(ED in)
.165 E(byte mode, or of an)100.8 603 Q 2.75(yA)-.165 G
(QM mechanisms with similar beha)-3.355 E 2.75(vior; we)-.22 F(don')2.75
E 2.75(tk)-.198 G(no)-2.75 E 2.75(wt)-.275 G(he e)-2.75 E(xtent)-.165 E
(of the deplo)100.8 616 Q(yment of standard RED, or or an)-.11 E 2.75
(yo)-.165 G 2.75(ft)-2.75 G(he proposed A)-2.75 E(QM mechanisms.)-.605 E
(Flo)72 733 Q(yd/K)-.11 E 268.584(ohler Section)-.385 F 2.75(B.4. [P)
2.75 F(age 32])-.165 E 0 Cg EP
%%Page: 33 33
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Courier@0 SF 6.6
(<-----S)226.2 85 S(end Rates in Kbps - - - - >)-6.6 E 46.2(Web TCP)
166.8 98 R(\(512B seg\))6.6 E(TFRC-SP \(200B seg\))39.6 E 13.2
(Sessions DropRate)153.6 111 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)153.6 124 R 19.8(-------- --------)13.2 F
(--------)13.2 E 39.6(10 0.01)180 137 R 39.6(306.56 0.01)33 F(185.11)33
E 39.6(25 0.02)180 150 R 39.6(261.41 0.01)33 F(184.41)33 E 39.6(50 0.02)
180 163 R 39.6(185.07 0.01)33 F(184.54)33 E 39.6(100 0.04)173.4 176 R
39.6(59.25 0.03)39.6 F(181.58)33 E 39.6(200 0.08)173.4 189 R 39.6
(16.32 0.06)39.6 F(150.87)33 E 39.6(400 0.12)173.4 202 R 39.6
(11.53 0.10)39.6 F(98.10)39.6 E 39.6(800 0.25 5.85 0.15)173.4 215 R
(46.59)39.6 E 39.6(1600 0.32 1.43 0.22)166.8 228 R(19.40)39.6 E
(Table 21: Drop and Send Rates for Adaptive RED Queues in Byte Mode)
100.8 254 Q F0 -.88(Ta)100.8 276.6 S(ble 21 sho).88 E
(ws that when TFRC-SP with 240-byte pack)-.275 E
(ets competes with TCP with)-.11 E(552-byte pack)100.8 289.6 Q
(ets in a scenario with Adapti)-.11 E .33 -.165(ve R)-.275 H
(ED in byte mode, the TFRC-SP \215o).165 E(ws still)-.275 E(recei)100.8
302.6 Q .33 -.165(ve m)-.275 H(ore bandwidth that the TCP \215o).165 E
(ws, b)-.275 E(ut the le)-.22 E -.165(ve)-.275 G 2.75(lo).165 G 2.75(fu)
-2.75 G(nf)-2.75 E(airness is less se)-.11 E -.165(ve)-.275 G(re, and)
.165 E(the pack)100.8 315.6 Q(et drop rates of the TCP \215o)-.11 E
(ws is at most twice that of the TFRC-SP \215o)-.275 E 2.75(ws. That)
-.275 F(is,)2.75 E(the TFRC-SP \215o)100.8 328.6 Q(ws still recei)-.275
E .33 -.165(ve m)-.275 H(ore than their share of the bandwidth, b).165 E
(ut the TFRC-SP)-.22 E(congestion control is more ef)100.8 341.6 Q
(fecti)-.275 E .33 -.165(ve t)-.275 H(hat than in T).165 E(able 20 abo)
-.88 E -.165(ve)-.165 G(.).165 E/F2 11/Times-Bold@0 SF(C.)72 367.6 Q/F3
14/Times-Bold@0 SF -.35(Ap)5.5 G(pendix: Exploring P).35 E
(ossible Oscillations in the Loss Ev)-.28 E(ent Rate)-.14 E F0
(TFRC-SP estimates the loss interv)100.8 384.2 Q(al size dif)-.275 E
(ferently for short and long loss interv)-.275 E(als,)-.275 E
(counting only one loss e)100.8 397.2 Q -.165(ve)-.275 G
(nt for long loss interv).165 E(als, b)-.275 E(ut counting all pack)-.22
E(et losses as loss)-.11 E -2.365 -.275(ev e)100.8 410.2 T
(nts for the short loss interv).275 E 2.75(als. One)-.275 F
(question that has been raised is whether this can)2.75 E
(lead to oscillations in the a)100.8 423.2 Q -.165(ve)-.22 G
(rage loss e).165 E -.165(ve)-.275 G(nt rate in en).165 E
(vironments where there are man)-.44 E(y)-.165 E(pack)100.8 436.2 Q
(et drops in a single loss e)-.11 E -.165(ve)-.275 G(nt, and loss e).165
E -.165(ve)-.275 G(nts switch from short to long and vice).165 E -.165
(ve)100.8 449.2 S 2.75(rsa. As).165 F(an e)2.75 E
(xample, consider a loss interv)-.165 E(al with N pack)-.275 E
(ets, including N/4 losses.)-.11 E(If this)5.5 E(loss interv)100.8 462.2
Q(al is short \(at most tw)-.275 E 2.75(or)-.11 G
(ound-trip times\), the loss interv)-2.75 E(al length is measured as)
-.275 E 2.75(4p)100.8 475.2 S(ack)-2.75 E 2.75(ets. If)-.11 F
(this loss interv)2.75 E(al is long, then the loss interv)-.275 E
(al length is measured as N)-.275 E(pack)100.8 488.2 Q(ets.)-.11 E
(If the loss interv)100.8 504.8 Q
(al switching from short to long and back leads to se)-.275 E -.165(ve)
-.275 G(re oscillations in the).165 E -2.475 -.22(av e)100.8 517.8 T
(rage pack).22 E(et drop rate, and therefore in the allo)-.11 E
(wed sending rate, one solution w)-.275 E(ould be)-.11 E(to ha)100.8
530.8 Q .33 -.165(ve a m)-.22 H
(ore gradual transition between the calculation of the loss interv).165
E(al length for)-.275 E(short and long loss interv)100.8 543.8 Q 2.75
(als. F)-.275 F(or e)-.165 E(xample, one possibility w)-.165 E
(ould be to use all of the pack)-.11 E(et)-.11 E
(losses for a loss interv)100.8 556.8 Q
(al of one round-trip time in calculating the loss interv)-.275 E
(al length, to use)-.275 E(2/3-rds of the pack)100.8 569.8 Q
(et losses from a loss interv)-.11 E(al of tw)-.275 E 2.75(or)-.11 G
(ound-trip times, to use 1/3-rd of the)-2.75 E(pack)100.8 582.8 Q
(et losses from a loss interv)-.11 E
(al of three round-trip time, and to use only one pack)-.275 E(et loss)
-.11 E(from a loss interv)100.8 595.8 Q
(al of four or more round-trip times.)-.275 E
(This more gradual mechanism w)5.5 E(ould)-.11 E(gi)100.8 608.8 Q .33
-.165(ve a g)-.275 H
(radual transition to counting all losses for a loss interv).165 E
(al of only one round-trip)-.275 E(time, and counting only one loss e)
100.8 621.8 Q -.165(ve)-.275 G(nt for a loss interv).165 E
(al of four or more round-trip times.)-.275 E(Ho)100.8 638.4 Q(we)-.275
E -.165(ve)-.275 G .88 -.44(r, o).165 H(ur simulations so f).44 E(ar ha)
-.11 E .33 -.165(ve n)-.22 H(ot sho).165 E(wn a great dif)-.275 E
(ference in oscillations in the)-.275 E(estimate loss e)100.8 651.4 Q
-.165(ve)-.275 G(nt rate between the def).165 E
(ault mechanism for estimating the loss interv)-.11 E(al)-.275 E
(length for short loss interv)100.8 664.4 Q
(al and the gradual mechanism described abo)-.275 E -.165(ve)-.165 G 5.5
(.S).165 G(imulation)-5.5 E(scripts are a)100.8 677.4 Q -.275(va)-.22 G
(ilable from [V).275 E 2.75(OIPSIMS]. Therefore,)-.44 F(for no)2.75 E
2.75(ww)-.275 G 2.75(ea)-2.75 G(re staying with the simple)-2.75 E(def)
100.8 690.4 Q(ault mechanism for estimating the loss e)-.11 E -.165(ve)
-.275 G(nt rate for short loss interv).165 E(als described in)-.275 E
(Flo)72 733 Q(yd/K)-.11 E 276.834(ohler Section)-.385 F 2.75(C. [P)2.75
F(age 33])-.165 E 0 Cg EP
%%Page: 34 34
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(this document.)100.8 85 Q/F1
11/Times-Bold@0 SF -.22(D.)72 111 S/F2 14/Times-Bold@0 SF -.35(Ap)5.72 G
(pendix: A Discussion of P).35 E(ack)-.14 E(et Size and P)-.14 E(ack)
-.14 E(et Dr)-.14 E(opping)-.252 E F0(The table belo)100.8 127.6 Q 2.75
(wg)-.275 G -2.365 -.275(iv e)-2.75 H 2.75(sag).275 G
(eneral summary of the relati)-2.75 E .33 -.165(ve a)-.275 H(dv).165 E
(antages of pack)-.275 E(et-dropping)-.11 E(beha)100.8 140.6 Q
(vior at routers independent of pack)-.22 E(et size, v)-.11 E
(ersus pack)-.165 E(et dropping beha)-.11 E(vior where)-.22 E(lar)100.8
153.6 Q(ge pack)-.198 E(ets are more lik)-.11 E
(ely to be dropped than small ones.)-.11 E/F3 11/Courier@0 SF
(Advantages of Packet Dropping Independent of Packet Size:)100.8 172.6 Q
(---------------------------------------------------------)100.8 185.6 Q
6.6(1. Adds)100.8 198.6 R
(another incentive for end nodes to use large packets.)6.6 E 6.6
(2. Matches)100.8 224.6 R
(an environment with a limitation in pps rather than)6.6 E(bps.)127.2
237.6 Q(---------------------------------------------------------)100.8
250.6 Q(Advantages of Packet Dropping as a Function of Packet Size:)
100.8 276.6 Q(---------------------------------------------------------)
100.8 289.6 Q 6.6(1. Small)100.8 302.6 R
(control packets are less likely to be dropped than are)6.6 E
(large data packets, improving TCP performance.)127.2 315.6 Q 6.6
(2. Matches)100.8 341.6 R
(an environment with a limitation in bps rather than)6.6 E(pps.)127.2
354.6 Q 6.6(3. Reduces)100.8 380.6 R
(the penalty of TCP and other transport protocols)6.6 E
(against flows with small packets \(where the allowed sending)127.2
393.6 Q(rate is roughly a linear function of packet size\).)127.2 406.6
Q 6.6(4. A)100.8 432.6 R
(queue limited in bytes is better than a queue limited in)6.6 E
(packets for matching the worst-case queue size to the)127.2 445.6 Q
(worst-case queueing delay in seconds.)127.2 458.6 Q
(---------------------------------------------------------)100.8 471.6 Q
F2(Normati)72 503.6 Q .28 -.14(ve R)-.14 H(efer).14 E(ences)-.252 E F0
29.634([RFC2119] S.)113.1 520.2 R(Bradner)2.75 E 2.75(.K)-.605 G .33
-.165(ey W)-3.025 H(ords F)-.715 E
(or Use in RFCs to Indicate Requirement)-.165 E(Le)195.6 533.2 Q -.165
(ve)-.275 G(ls. RFC 2119.).165 E 29.634([RFC3448] M.)113.1 549.8 R
(Handle)2.75 E 1.43 -.715(y, S)-.165 H 2.75(.F).715 G(lo)-2.75 E
(yd, J. P)-.11 E(adh)-.165 E(ye, and J. W)-.055 E(idmer)-.44 E 2.75(,T)
-.44 G(CP Friendly Rate)-2.75 E
(Control \(TFRC\): Protocol Speci\214cation, RFC 3448, Proposed)195.6
562.8 Q(Standard, January 2003.)195.6 575.8 Q F2(Inf)72 601.8 Q(ormati)
-.35 E .28 -.14(ve R)-.14 H(efer).14 E(ences)-.252 E F0 40.645
([CCID3] S.)113.1 618.4 R(Flo)2.75 E(yd, E. K)-.11 E(ohler)-.385 E 2.75
(,a)-.44 G(nd J. P)-2.75 E(adh)-.165 E 2.75(ye. Pro\214le)-.055 F
(for DCCP Congestion)2.75 E(Control ID 3: TFRC Congestion Control.)195.6
631.4 Q(draft-ietf-dccp-)5.5 E(ccid3-11.txt, w)195.6 644.4 Q
(ork in progress, March 2005.)-.11 E 43.692([DCCP] E.)113.1 661 R -.385
(Ko)2.75 G(hler).385 E 2.75(,M)-.44 G 2.75(.H)-2.75 G(andle)-2.75 E 1.43
-.715(y, a)-.165 H(nd S. Flo).715 E 2.75(yd. Datagram)-.11 F
(Congestion Control)2.75 E(Protocol, draft-ietf-dccp-spec-13.txt, w)
195.6 674 Q(ork in progress, December)-.11 E(2005.)195.6 687 Q(Flo)72
733 Q(yd/K)-.11 E 328.171(ohler [P)-.385 F(age 34])-.165 E 0 Cg EP
%%Page: 35 35
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E 46.761([EA03] W)113.1 85 R
2.75(.E)-1.012 G(ddy and M. Allman.)-2.75 E 2.75(AC)5.5 G
(omparison of RED')-2.75 E 2.75(sB)-.605 G(yte and P)-2.75 E(ack)-.165 E
(et)-.11 E(Modes, Computer Netw)195.6 98 Q(orks, 42\(2\), June 2003.)
-.11 E 55.308([P04] T)113.1 114.6 R 2.75(.P)-.814 G
(helan, TFRC with Self-Limiting Sources, October 2004.)-2.75 E(URL)5.5 E
("http://www)195.6 127.6 Q(.phelan-4.com/dccp/".)-.715 E 29.634
([RFC3168] K.)113.1 144.2 R(Ramakrishnan, S. Flo)2.75 E
(yd, and D. Black, The Addition of Explicit)-.11 E
(Congestion Noti\214cation \(ECN\) to IP)195.6 157.2 Q 2.75(,R)-1.221 G
(FC 3168, September 2001.)-2.75 E 29.634([RFC3714] S.)113.1 173.8 R(Flo)
2.75 E(yd and J. K)-.11 E(empf, Editors.)-.275 E(IAB Concerns Re)5.5 E
-.055(ga)-.165 G(rding).055 E(Congestion Control for V)195.6 186.8 Q
(oice T)-1.419 E(raf)-.385 E(\214c in the Internet.)-.275 E(RFC 3714.)
5.5 E 16.797([RFC3448bis] M.)113.1 203.4 R(Handle)2.75 E 1.43 -.715
(y, S)-.165 H 2.75(.F).715 G(lo)-2.75 E(yd, J. P)-.11 E(adh)-.165 E
(ye, and J. W)-.055 E(idmer)-.44 E 2.75(,T)-.44 G(CP Friendly Rate)-2.75
E(Control \(TFRC\): Protocol Speci\214cation, internet draft draft-ietf\
-dccp-)195.6 216.4 Q(rfc3448bis-00.txt, w)195.6 229.4 Q
(ork in progress, October 2006.)-.11 E 55.308([S05] Peter)113.1 246 R
(Sholander)2.75 E 2.75(,p)-.44 G(ri)-2.75 E -.275(va)-.275 G
(te communications, August 2005.).275 E(Citation for)5.5 E(ackno)195.6
259 Q(wledgement purposes only)-.275 E(.)-.715 E 53.482([V00] P)113.1
275.6 R 2.75(.V)-1.221 G 2.75(asallo. V)-3.971 F(ariable P)-1.221 E(ack)
-.165 E(et Size Equation-Based Congestion Control.)-.11 E(ICSI T)195.6
288.6 Q(echnical Report TR-00-008, April 2000.)-.77 E(URL)5.5 E
("http://www)195.6 301.6 Q(.icsi.berk)-.715 E(ele)-.11 E -.715(y.)-.165
G(edu/techreports/).715 E(2000.abstracts/tr)195.6 314.6 Q
(-00-008.html".)-.22 E([V)113.1 331.2 Q 21.527(OIPSIMS] W)-.44 F
(eb page "http://www)-.88 E(.icir)-.715 E(.or)-.605 E(g/tfrc/v)-.198 E
(oipsims.html".)-.22 E 36.982([WBL04] J.)113.1 347.8 R -.44(Wi)2.75 G
(dmer).44 E 2.75(,C)-.44 G 2.75(.B)-2.75 G(outremans, and Jean-Yv)-2.75
E(es Le Boudec.)-.165 E(Congestion)5.5 E(Control for Flo)195.6 360.8 Q
(ws with V)-.275 E(ariable P)-1.221 E(ack)-.165 E(et Size, A)-.11 E
(CM CCR, 34\(2\), 2004.)-.44 E/F1 14/Times-Bold@0 SF -.7(Au)72 386.8 S
(thors' Addr).7 E(esses)-.252 E F0(Sally Flo)100.8 403.4 Q(yd <\215o)
-.11 E(yd@icir)-.11 E(.or)-.605 E(g>)-.198 E
(ICSI Center for Internet Research)100.8 416.4 Q
(1947 Center Street, Suite 600)100.8 429.4 Q(Berk)100.8 442.4 Q(ele)-.11
E 1.43 -.715(y, C)-.165 H 2.75(A9).715 G(4704)-2.75 E(USA)100.8 455.4 Q
(Eddie K)100.8 472 Q(ohler <k)-.385 E(ohler@cs.ucla.edu>)-.11 E
(4531C Boelter Hall)100.8 485 Q(UCLA Computer Science Department)100.8
498 Q(Los Angeles, CA 90095)100.8 511 Q(USA)100.8 524 Q F1
(Full Copyright Statement)72 550 Q F0(Cop)100.8 566.6 Q
(yright \(C\) The Internet Society \(2006\).)-.11 E
(This document is subject to the rights, licenses)5.5 E
(and restrictions contained in BCP 78, and e)100.8 579.6 Q
(xcept as set forth therein, the authors retain all)-.165 E
(their rights.)100.8 592.6 Q
(This document and the information contained herein are pro)100.8 609.2
Q(vided on an "AS IS" basis and)-.165 E(THE CONTRIB)100.8 622.2 Q(UT)
-.11 E(OR, THE ORGANIZA)-.198 E(TION HE/SHE REPRESENTS OR IS)-1.221 E
(SPONSORED BY \(IF ANY\), THE INTERNET SOCIETY AND THE INTERNET)100.8
635.2 Q(ENGINEERING T)100.8 648.2 Q(ASK FORCE DISCLAIM ALL W)-1.023 E
(ARRANTIES, EXPRESS OR)-1.32 E(IMPLIED, INCLUDING B)100.8 661.2 Q(UT NO)
-.11 E 2.75(TL)-.44 G(IMITED T)-2.75 E 2.75(OA)-.198 G(NY W)-2.75 E
(ARRANTY THA)-1.32 E 2.75(TT)-1.221 G(HE USE)-2.75 E(OF THE INFORMA)
100.8 674.2 Q(TION HEREIN WILL NO)-1.221 E 2.75(TI)-.44 G
(NFRINGE ANY RIGHTS OR ANY)-2.75 E(IMPLIED W)100.8 687.2 Q
(ARRANTIES OF MERCHANT)-1.32 E(ABILITY OR FITNESS FOR A)-1.023 E(Flo)72
733 Q(yd/K)-.11 E 328.171(ohler [P)-.385 F(age 35])-.165 E 0 Cg EP
%%Page: 36 36
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E -1.012(PA)100.8 85 S -.66(RT)
1.012 G(ICULAR PURPOSE.).66 E/F1 14/Times-Bold@0 SF(Intellectual Pr)72
111 Q(operty)-.252 E F0(The IETF tak)100.8 127.6 Q(es no position re)
-.11 E -.055(ga)-.165 G(rding the v).055 E(alidity or scope of an)-.275
E 2.75(yI)-.165 G(ntellectual Property)-2.75 E(Rights or other rights t\
hat might be claimed to pertain to the implementation or use of the)
100.8 140.6 Q(technology described in this document or the e)100.8 153.6
Q(xtent to which an)-.165 E 2.75(yl)-.165 G(icense under such rights)
-2.75 E(might or might not be a)100.8 166.6 Q -.275(va)-.22 G
(ilable; nor does it represent that it has made an).275 E 2.75(yi)-.165
G(ndependent)-2.75 E(ef)100.8 179.6 Q(fort to identify an)-.275 E 2.75
(ys)-.165 G(uch rights.)-2.75 E
(Information on the procedures with respect to rights in)5.5 E
(RFC documents can be found in BCP 78 and BCP 79.)100.8 192.6 Q
(Copies of IPR disclosures made to the IETF Secretariat and an)100.8
209.2 Q 2.75(ya)-.165 G(ssurances of licenses to)-2.75 E(be made a)100.8
222.2 Q -.275(va)-.22 G(ilable, or the result of an attempt made to obt\
ain a general license or).275 E(permission for the use of such propriet\
ary rights by implementers or users of this)100.8 235.2 Q(speci\214cati\
on can be obtained from the IETF on-line IPR repository at)100.8 248.2 Q
(http://www)100.8 261.2 Q(.ietf.or)-.715 E(g/ipr)-.198 E(.)-.605 E
(The IETF in)100.8 277.8 Q(vites an)-.44 E 2.75(yi)-.165 G
(nterested party to bring to its attention an)-2.75 E 2.75(yc)-.165 G
(op)-2.75 E(yrights, patents or)-.11 E
(patent applications, or other proprietary rights that may co)100.8
290.8 Q -.165(ve)-.165 G 2.75(rt).165 G(echnology that may be)-2.75 E
(required to implement this standard.)100.8 303.8 Q
(Please address the information to the IETF at ietf-)5.5 E(ipr@ietf.or)
100.8 316.8 Q(g.)-.198 E(Flo)72 733 Q(yd/K)-.11 E 328.171(ohler [P)-.385
F(age 36])-.165 E 0 Cg EP
%%Trailer
end
%%EOF